@proceedings{ANNGA93, title = {Proceedings of the International Conference on Artificial Neural Networks and Genetic Algorithms}, year = {1993}, month = {April 14-16}, editor = {Rudolf F. Albrecht and Colin R. Reeves and Nigel C. Steele}, publisher = {Springer-Verlag} } @proceedings{AlifeI, title = {Artificial Life: the Proceedings of an Interdisciplinary Workshop on the Synthesis and Simulation of Living Systems}, year = {1989}, month = {September}, editor = {Christopher G. Langton}, publisher = {Addison-Wesley}, address = {Redwood City, CA}, note = {Workshop held September, 1987 in Los Alamos, New Mexico} } @proceedings{AlifeII, title = {Artificial Life II: Proceedings of the Workshop on Artificial Life}, year = {1992}, editor = {Christopher G. Langton and Charles Taylor and J. Doyne Farmer and Steen Rasmussen}, publisher = {Addison-Wesley}, address = {Redwood City, Calif.}, note = {Workshop held February, 1990 in Santa Fe, New Mexico} } @proceedings{AlifeIII, title = {Artificial Life III: Proceedings of the Workshop on Artificial Life}, year = {1994}, editor = {Christopher G. Langton}, publisher = {Addison-Wesley}, address = {Reading, MA}, note = {Workshop held June, 1992 in Santa Fe, New Mexico} } @proceedings{COGANN92, title = {International Workshop on Combinations of Genetic Algorithms and Neural Networks: COGANN-92}, year = {1992}, editor = {L. D. Whitley and J. D. Schaffer}, publisher = {IEEE Computer Society Press}, address = {Los Alamiitos, California} } @proceedings{FOGA1, title = {Proceedings of the Workshop on Foundations of Genetic Algorithms}, year = {1991}, editor = {Gregory J. E. Rawlins}, publisher = {Morgan Kaufmann}, address = {San Mateo, California} } @proceedings{FOGA2, title = {Proceedings of the Workshop on Foundations of Genetic Algorithms}, year = {1993}, editor = {Darrell L. Whitley}, publisher = {Morgan Kaufmann}, address = {San Mateo, California}, note = {The second workshop on Foundations of Genetic Algorithms (FOGA) was held July 26-29, 1992 in Vail, Colorado} } @proceedings{ICGA85, title = {Proceedings of the First International Conference on Genetic Algorithms and their Applications}, year = {1985}, month = {July 24-26}, editor = {John J. Grefenstette}, publisher = {Lawrence Erlbaum Associates}, address = {Pittsburgh, Pa} } @proceedings{ICGA87, title = {Proceedings of the Second International Conference on Genetic Algorithms and their Applications}, organization = {Massachusetts Institute of Technology, Cambridge, MA}, year = {1987}, month = {July 28-31}, editor = {John J. Grefenstette}, publisher = {Lawrence Erlbaum Associates}, address = {Hillsdale, New Jersey} } @proceedings{ICGA89, title = {Proceedings of the Third International Conference on Genetic Algorithms}, organization = {George Mason University}, year = {1989}, month = {June 4-7}, editor = {J. David Schaffer}, publisher = {Morgan Kaufmann}, address = {San Mateo, California} } @proceedings{ICGA91, title = {Proceedings of the Fourth International Conference on Genetic Algorithms}, organization = {University of California, San Diego}, year = {1991}, month = {July 13-16}, editor = {Richard K. Belew and Lashon B. Booker}, publisher = {Morgan Kaufmann}, address = {San Mateo, CA} } @proceedings{ICGA93, title = {Proceedings of the Fifth International Conference on Genetic Algorithms}, organization = {University of Illinois at Urbana Champaign}, year = {1993}, month = {July 17-21}, editor = {Stephanie Forrest}, publisher = {Morgan Kaufmann}, address = {San Mateo, CA} } @proceedings{PPSN91, title = {Proceedings of the First Conference on Parallel Problem Solving from Nature}, year = {1991}, month = {October 1-3}, editor = {Hans-Paul Schwefel and Reinhart M{\"a}nner}, publisher = {Springer-Verlag}, address = {Dortmund, Germany}, volume = {496}, series = {Lecture Notes in Computer Science} } @proceedings{PPSN92, title = {Proceedings of the Second Conference on Parallel Problem Solving from Nature, Brussels, Belgium}, year = {1992}, month = {September 28-30}, editor = {Reinhart M{\"a}nner and Bernhard Manderick}, publisher = {Elsevier}, address = {Amsterdam} } @conference{Ackley85, key = {genetic algorithm, boltzmann, connectionism, cogann ref}, author = {David H. Ackley}, title = {A Connectionist Algorithm for Genetic Search}, booktitle = {Proceedings of the First International Conference on Genetic Algorithms and their Applications}, year = {1985}, editor = {John. J Grefenstette}, publisher = {Lawrence Erlbaum Associates}, address = {Hillsdale, New Jersey}, pages = {121-135}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @book{Ackley87, key = {connectionism, genetic algorithm, sigh, stochastic iterated, cogann ref}, author = {David H. Ackley}, title = {A Connectionist Machine for Genetic Hillclimbing}, year = {1987}, publisher = {Kluwer Academic Publishers}, address = {Boston, MA}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Ackley92, author = {David H. Ackley and Michael L. Littman}, title = {Interactions between Learning and Evolution}, booktitle = {Artificial Life II}, year = {1992}, editor = {Christopher G. Langton and Charles Taylor and J. Doyne Farmer and Steen Rasmussen}, publisher = {Addison}, pages = {487-509}, annote = {connectionism genetic algorithm neighborhood mate selection, cogann ref animat}, topology = {feed-forward}, network = { }, encoding = {direct}, evolves = {parameters, connectivity}, applications = {simulated world} } @inproceedings{Alba93, key = {genetic algorithms connectionism neural networks cogann}, author = {E. Alba and J.F. Aldana and J.M. Troya}, title = {Genetic Algorithms as Heuristics for Optimizing ANN Design}, booktitle = {Proceedings of the International Conference on Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {683-690}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @article{Angeline94, key = {genetic algorithms connectionism neural networks cogann programming}, author = {Peter J. Angeline and Gregory M. Saunders and Jordan B. Pollack}, title = {An Evolutionary Algorithm that Constructs Recurrent Neural Networks}, journal = {IEEE Transactions on Neural Networks}, year = {1994}, volume = {5}, pages = {54-64}, topology = {recurrent}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @article{Ankenbrandt90, key = {Connectionism, fuzzy logic, pattern recognition}, author = {C. A. Ankenbrandt and B. P. Buckles and F. E. Petry}, title = {Scene Recognition using Genetic Algorithms with Semantic Nets}, journal = {Pattern Recognition Letters}, year = {1990}, month = {April}, volume = {11}, pages = {285-293}, publisher = {North-Holland}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {image recognition} } @inproceedings{Arena93, key = {genetic algorithms connectionism neural networks cogann}, author = {P. Arena and R. Caponetto and I. Fortuna and M. G. Xibilia}, title = {MLP Optimal Topology via Genetic Algorithms}, booktitle = {Proceedings of the International Conference on Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {670-674}, topology = {multi-layered}, network = {multi-layer perceptron}, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Austin91, key = {algorithms connectionism, cogann ref}, author = {Scott Austin}, title = {Genetic Neurosynthesis}, booktitle = {Proceedings of AIAA Aerospace VIII}, year = {1991}, month = {October}, address = {Baltimore, MD}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Ball90, key = {connectionism, cogann ref}, author = {N. Ball}, title = {Adaptive Signal Processing via Genetic Algorithms and Self-organizing Neural Networks}, booktitle = {Proceedings of the IEEE Workshop on Genetic Algorithms, Simulated Annealing and Neural Networks}, year = {1990}, address = {University of Glasgow, Scotland}, network = {self-organizing}, encoding = { }, evolves = { }, applications = {signal processing} } @inproceedings{Ball93, key = {genetic algorithms connectionism neural networks cogann}, author = {N. R. Ball}, title = {Towards the Development of Cognitive Maps in Classifier Systems}, booktitle = {Proceedings of the International Conference on Artificial Neural Nets and Genetic Algorithms}, year = {1993}, pages = {712-718}, topology = { }, network = { }, encoding = {indirect, classifier systems}, evolves = { }, applications = {cognitive maps} } @article{Beer92, key = {genetic algorithms GENESIS, connectionism}, author = {Randall D. Beer and John C. Gallagher}, title = {Evolving Dynamical Neural Networks for Adaptive Behavior}, journal = {Adaptive Behavior}, year = {1992}, volume = {1}, number = {1}, topology = {recurrent}, network = { }, encoding = { }, evolves = {connectivity}, applications = {animat controller} } @conference{Belew89, key = {hybrid learning, connectionism, cogann ref}, author = {Richard K. Belew}, title = {When Both Individuals and Populations Search: Adding Simple Learning to the Genetic Algorithm}, booktitle = {Proceedings of the Third International Conference on Genetic Algorithms}, organization = {ICGA89}, year = {1989}, editor = {Schaffer, J. D}, publisher = {Morgan Kaufmann}, pages = {34-41}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Belew89a, key = {connectionism, genetic algorithms, cogann ref}, author = {Richard K. Belew}, title = {Evolution, Learning and Culture: Computational Metaphors for Adaptive Algorithms}, institution = {University of California at San Deigo}, year = {1989}, month = {September}, address = {La Jolla, CA}, type = {CSE Technical Report CS89-156}, publisher = {University of California at San Deigo}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Belew89b, key = {connectionism, cogann ref}, author = {Richard K. Belew and John McInerney}, title = {Using the Genetic Algorithm to Wire Feed-forward Networks}, institution = {University of California, San Diego}, year = {1989}, month = {May}, address = {La Jolla, CA}, type = {Technical abstract}, note = {Submitted to Neural Information Processing Systems 1989}, publisher = {Computer Science \& Engineering Dept., University of California at San Deigo}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @techreport{Belew90, key = {neural nets, cogann ref}, author = {Richard K. Belew and John McInerney and Nicol N. Schraudolph}, title = {Evolving Networks: Using Genetic Algorithms with Connectionist Learning}, institution = {University of California at San Diego}, year = {1990}, month = {June}, address = {La Jolla, CA}, type = {CSE Technical Report CS90-174}, publisher = {University of California at San Diego}, topology = {feed-forward}, network = { }, encoding = {direct, developmental}, evolves = {parameters}, applications = { } } @inproceedings{Bengio91, key = {genetic algorithms connectionism neural networks cogann}, author = {Yoshua Bengio and Samy Bengio and Jocelyn Cloutier}, title = {Learning a synaptic learning rule}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1991}, pages = {969}, abstract = {ABSTRACT Summary form only given, as follows. The Authors discuss original approach to neural modeling based on the idea of searching, with learning methods, for a synaptic learning rule which is biologically plausible and yields networks that are able to learn to perform difficult tasks. The proposed method of automatically finding the learning rule relies on the idea of considering the synaptic modification rule as a parametric function. This function has local inputs and is the same in many neurons. The parameters that define this function can be estimated with known learning methods. For this optimization, particular attention is given to gradient descent and genetic algorithms. In both cases, estimation of this function consists of a joint global optimization of the synaptic modification function and the networks that are learning to perform some tasks. Both network architecture and the learning function can be designed within constraints derived from biological knowledge.}, topology = { }, network = { }, encoding = { }, evolves = {learning rule}, applications = { } } @article{Bergman88, author = {Aviv Bergman}, title = {Variation and Selection: An Evolutionary Model of Learning in Neural Networks}, journal = {Neural Networks}, year = {1988}, volume = {1}, number = {1}, pages = {75-}, network = { }, encoding = { }, applications = { } } @inproceedings{Bergman89, key = {genetic algorithms, connectionism recurrent neural networks, cogann ref}, author = {Aviv Bergman}, title = {Self-Organization by Simulated Evolution}, booktitle = {Lectures in Complex Systems: Proceedings of the 1989 Complex Systems Summer School}, year = {1989}, editor = {E. Jen}, address = {Santa Fe}, network = {self-organizing}, encoding = { }, evolves = { }, applications = { } } @inproceedings{Bergman87, key = {genetic algorithms, connectionism, evolution, cogann ref}, author = {Aviv Bergman and Michel Kerszberg}, title = {Breeding Intelligent Automata}, booktitle = {Proceedings of IEEE Conference on Neural Networks}, year = {1987}, month = {June 21-24}, address = {San Diego, CA}, pages = {63-70}, volume = {II}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Bessiere92, key = {genetic algorithms connectionism neural networks cogann}, author = {P. Bessiere}, title = {Genetic Algorithms Applied to Formal Neural Networks: Parallel Genetic Implementation of a Boltzmann Machine and Associated Robotic Experimentations}, booktitle = {Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life}, year = {1992}, editor = {F.J. Varela and P. Bourgine}, publisher = {MIT Press}, address = {Cambridge, MA, USA}, pages = {310-314}, abstract = {ABSTRACT Describes a possible application of computing techniques inspired by natural life mechanisms to an artificial life creature, namely a small mobile robot, called KitBorg. Probabilistic inference suggests that any cognitive problem may be split in two optimization problems. The first one called the dynamic inference problem is an abstraction of learning, the second one, namely, the static inference problem, being a mathematical metaphor of pattern association. Other optimization technics should be considered in that context and especially genetic algorithms. The purpose of this paper is to describe the state of the art of the investigations which the Author is "akin" about that question using a parallel genetic algorithm. The author first "ecall" the principles of probabilistic inference, then he presents briefly the parallel genetic algorithm and the ways it is used to deal with both optimization problems, to finally conclude about ongoing robotic experimentations and future planned extensions.}, network = {boltzmann}, encoding = { }, evolves = { }, applications = {robot controller} } @inproceedings{Bishop93, key = {genetic algorithms connectionism neural networks cogann application paint industry}, author = {J.M. Bishop and M.J. Bushnell and A. Usher and S. Westland}, title = {Genetic Optimization of Neural Network Architectures for Colour Recipe Prediction}, booktitle = {Proceedings of the International Conference on Artificial Neural Networks and Genetic Algorithms}, year = {1993}, pages = {719-725}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = {optimization} } @article{Bornholdt92, key = {connectionism, cogann ref}, author = {Stephan Bornholdt and Dirk Graudenz}, title = {General Assymetric Neural Networks and Structure Design by Genetic Algorithms}, journal = {Neural Networks}, year = {1992}, volume = {5}, number = {2}, pages = {327-334}, note = {DESSY 91-046, Deutsches Electronen-Synchrotron, Hamburg, Germany, MAY 1991}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @techreport{Bornholdt93, key = {connectionism, cogann}, author = {Stephan Bornholdt and Dirk Graudenz}, title = {General Assymetric Neural Networks and Structure Design by Genetic Algorithms: A Learning Rule for Temporal Patterns}, institution = {Lawrence Berkeley Laboratory, University of California}, year = {1993}, month = {July}, address = {Berkeley, CA}, type = {HD-THEP-93-26 LBL-34384}, publisher = {Lawrence Berkeley Laboratory, University of California}, abstract = {ABSTRACT A learning algorithm based on genetic algorithms for asymmetric neural networks with an arbitrary structure is presented. It is suited for the learning of temporal patterns and leads to stable neural networks with feedback.}, topology = {general, recurrent}, network = { }, encoding = { }, evolves = {connectivity}, applications = {temporal pattern recognition} } @article{Brassinne93, key = {genetic algorithms connectionism neural networks cogann}, author = {P. de la Brassinne}, title = {Genetic Algorithms and Learning of Neural Networks}, journal = {Bulletin Scientifique de l'Association des Ingenieurs Electriciens sortis de l'Institut Electrotechnique Montefiore}, year = {1993}, volume = {106}, number = {1}, pages = {41-,58}, abstract = {ABSTRACT The Author sought to apply genetic algorithms to two concrete industrial problems which caused trouble to classical optimization techniques (they were usually trapped into local minima), without positive results. One of the reasons was that the solutions among the population were too close to one another too early in the search process. Another was the unsuitability of the operators employed to create new solutions for the neural network optimization problem. Attempts at application to control problems, where backpropagation could not be used, yielded disappointing results except for very simple problems such as the inverted pendulum. An explanation of these findings is suggested.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {optimization} } @inproceedings{Braun93, key = {genetic algorithms connectionism neural networks cogann}, author = {H. Braun and J. Weisbrod}, title = {Evolving Neural Feedforward Networks}, booktitle = {Proceedings of the International Conference on Artificial Neural Networks and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C}, publisher = {Springer-Verlag}, pages = {25-32}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @article{Brill92, key = {genetic algorithms, connectionism, cogann ref}, author = {F.Z. Brill and D.E. Brown and W.N. Martin}, title = {Fast Genetic Selection of Features for Neural Network Classifiers}, journal = {IEEE Transactions on Neural Networks}, year = {1992}, month = {March}, volume = {3}, number = {2}, pages = {324-328}, abstract = {ABSTRACT - The task of classifiers is to determine the appropriate class name when presented with a sample from one of several classes. In forming the sample to present to the classifier, there may be a large number of measurements one can make. Feature selection addresses the problem of determining which of these measurements are the most useful for determining the pattern's class. In this paper, we describe experiments using a genetic algorithm for feature selection in the context of neural network classifiers, specifically, counterpropagation networks. We present two novel techniques in our application of genetic algorithms. First, we configure our genetic algorithm to use an approximate evaluation in order to reduce significantly the computation required. In particular, though our desired classifiers are counterpropagation networks, we use a nearest-neighbor classifier to evaluate feature sets. We show that the features selected by this method are effective in the context of counterpropagation networks. Second, we propose a method we call training set sampling, in which only a portion of the training set is used on any given evaluation. Again, significant computational savings can be made by using this method, i.e., evaluations can be made over an order of magnitude faster. This method selects feature sets that are as good as and occasionally better for counterpropagation than those chosen by an evaluation that uses the entire training set.}, topology = {counterpropagation}, network = { }, encoding = { }, evolves = { }, applications = {pattern classification} } @article{Bukatova92, key = {genetic algorithms connectionism neural networks cogann}, author = {I. L. Bukatova}, title = {Evolutionary Computer}, journal = {Proceedings of the RNNS/IEEE Symposium on Neuroinformatics and Neurocomputers.}, year = {1992}, month = {October 7-10}, volume = {I}, pages = {467-477}, address = {Rostov-on-Don, Russia}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Calvin87, author = {Calvin, W.H.}, title = {The Brain as a Darwin Machine}, journal = {Nature}, year = {1988}, volume = {330}, pages = {33-43}, network = { }, encoding = { }, applications = { } } @techreport{Carugo91, key = {connectionism, backpropagation, cogann ref}, author = {Marcelo H. Carugo}, title = {Optimization of Parameters of a Neural Network, applied to Document Recognition, using Genetic Algorithms}, institution = {N.V. Philips}, year = {1991}, address = {Eindhoven, The Netherlands}, type = {Nat. Lab. Technical Note No. 049/91}, publisher = {N.V. Philips}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {parameters}, applications = {image recognition} } @inproceedings{Caudell90, key = {genetic algorithms, neural networks, connectionism, constrained weights, implementation of neural networks, electro-optical systems, rms error minimization, convoluted error surfaces, problem: parity, parametric connectivity, cogann ref}, author = {Thomas P. Caudell}, title = {Parametric Connectivity: Feasibility of Learning in Constrained Weight Space}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1990}, pages = {667-675}, volume = {I}, abstract = {Uses constrained (linked) weights (ie, spread networks) trained via genetic search.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {parity} } @inproceedings{Caudell89, key = {neural networks, connectionism, constrained weights, implementation of neural networks, electro-optical systems, rms error minimization, convoluted error surfaces, problem: parity, parametric connectivity, cogann ref}, author = {Thomas P. Caudell and Charles P. Dolan}, title = {Parametric Connectivity: Training of Constrained Networks using Genetic Algorithms}, booktitle = {Proceedings of the Third International Conference on Genetic Algorithms}, year = {1989}, pages = {370-374}, abstract = {Uses constrained (linked) weights (ie, spread networks) trained via genetic search.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {parity} } @article{Caudill91, key = {genetic algorithm connectionism, cogann ref}, author = {Maureen Caudill}, title = {Evolutionary Neural Networks}, journal = {AI Expert}, year = {1991}, month = {March}, pages = {28-33}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Chalmers90, key = {cogann ref}, author = {David J. Chalmers}, title = {The Evolution of Learning: An Experiment in Genetic Connectionism}, booktitle = {Proceedings of the 1990 Connectionist Summer School}, year = {1990}, editor = {D.S. Touretsky and J.L. Elman and T.J Sejnowski and G.E. Hinton}, publisher = {Morgan Kaufmann}, pages = {81-90}, topology = { }, network = { }, encoding = { }, evolves = {learning rule}, applications = { } } @inproceedings{Chang91, key = {connectionism}, author = {E. Chang and R. Lippmann}, title = {Using Genetic Algorithms to Improve Pattern Classification Performance}, booktitle = {Neural Information Processing Systems -- NIPS 3}, year = {1991}, publisher = {Morgan Kaufmann}, pages = {797-803}, editors = {David Touretzky}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {pattern classification} } @inproceedings{Chen92, key = {genetic algorithms, connectionism}, author = {Qi Chen and W. A. Weigand}, title = {Neural Net Model of Batch Processes and Optimization Based on an Extended Genetic Algorithm}, booktitle = {Proceedings of the International Joint Conferenc on Neural Networks}, year = {1992}, pages = {IV-519 - IV-524}, abstract = {ABSTRACT This paper investigates the use of neural network for modeling the batch processes. The consideration of the dynamics of batch processes, a cascade neural network which is the combination of BPN and Euler's numerical integration method, is successfully used to model of batch processes. In terms of this neural network model, an extended genetic algorithm is adopted to generate the optimal trajectory for improving the desired process performance. The genetic algorithm is a general methodology for searching a solution space in a manner analogous to the natural selection procedure in biological evolution. With the motivation of modern genetic techonology, the rule-inducer genetic algorithm is proposed for dynamic optimization of batch processes. The simulation study of a typical biochemical process shows this neural network modeling technique has a good generalization of the batch process and the extended real-value genetic algorithm has a good capability to solve the complicated dynamical optimization problems.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {optimization} } @inproceedings{Chu93, key = {Connectionism, Genetic Algorithms, cogann}, author = {C. H. Chu and C. R. Chow}, title = {A Genetic Algorithm Approach to Supervised Learning for Multilayered Networks}, booktitle = {World Congress on Neural Networks}, year = {1993}, pages = {IV744 - IV747}, abstract = {ABSTRACT A neural network learning algorithm based on genetic algorithms (GAs) for multilayered networks is described. The present method does not require that the input-output pairs for each layer to be known "a priori", since all modules are trained concurrently. For an N-module system, N separate pools of chromosomes are maintained and updated. The algorithm is tested using the 4-bit parity problem and a classification problem. Experiment results are presented and discussed.}, topology = {feed-forward}, network = { }, encoding = { }, evolves = { }, applications = {4-parity, classification} } @inproceedings{Collins90, key = {cogann ref, genetic algorithms, connectionism}, author = {R. Collins and D. Jefferson}, title = {An Artificial Neural Network Representation for Artificial Organisms}, booktitle = {Proceedings of the Conference on Parallel Problem Solving from Nature}, year = {1990}, pages = {259-263}, topology = {general}, network = { }, encoding = {direct}, evolves = {parameters}, applications = {simulated world} } @incollection{Compiani89, author = {Compiani, M. and Montanari D. and Serra R. and Valastro G.}, title = {Classifier Systems and Neural Networks}, booktitle = {Parallel Architectures and Neural Networks}, year = {1989}, editor = {Caianiello E.R.}, publisher = {World Scientific Press, Singapore}, pages = {33-43}, network = { }, encoding = {indirect, classifier systems}, applications = { } } @inproceedings{Das92, key = {genetic algorithms, connectionism, neural networks}, author = {Rajarshi Das and Darrell Whitley}, title = {Genetic Sparse Distributed Memories.}, booktitle = {Proceedings of the Conference on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, pages = {97-107}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Dasgupta92, key = {genetic algorithms, connectionism, neural networks}, author = {Dipankar Dasgupta and Douglas McGregor}, title = {Designing Application-Specific Neural Networks using the Structured Genetic Algorithm.}, booktitle = {Proceedings of the International Conference on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, pages = {87-96}, topology = {feed-forward}, network = { }, encoding = {direct}, evolves = {parameters}, applications = {xor, 4-2-4 encoder-decoder} } @inproceedings{Davis88, key = {connectionism, neural networks, formal equivalence, classifier systems, mapping networks to classifiers genetic algorithms}, author = {Lawrence Davis}, title = {Mapping Classifier Systems into Neural Networks}, booktitle = {Proceedings of the Workshop on Neural Information Processing Systems 1}, year = {1988}, pages = {49-56}, topology = { }, network = { }, encoding = {indirect, classifier systems}, evolves = { }, applications = { } } @conference{Davis89, key = {novel operators, adaptive parameter optimization, neural networks, connectionism, cogann ref}, author = {Lawrence Davis}, title = {Adapting Operator Probabilities in Genetic Algorithms}, booktitle = {Proceedings of the Third International Conference on Genetic Algorithms}, year = {1989}, pages = {61-69}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Davis89a, author = {Lawrence Davis}, title = {Mapping Neural Networks into Classifier Systems}, booktitle = {Proceedings of the 3rd International Conference on Genetic Algorithms}, year = {1989}, pages = {375-378}, network = { }, applications = { } } @article{DeRouin92, key = {genetic algorithms connectionism neural networks cogann}, author = {E. DeRouin and J. Brown}, title = {Alternative Learning Methods for Training Neural Network Classifiers}, journal = {Proceedings of the SPIE - The International Society for Optical Engineering}, year = {1992}, volume = {1710, pt.1}, pages = {II-474-II-483}, abstract = {ABSTRACT Neural networks have proven very useful in the field of pattern classification by mapping input patterns into one of several categories. Rather than being specifically programmed, backpropagation networks (BPNs) 'learn' this mapping by exposure to a training set, a collection of input pattern samples matched with their corresponding output classification. The proper construction of this training set is crucial to successful training of a BPN. One of the criteria to be met for proper construction of a training set is that each of the classes must be adequately represented. A class that is represented less often in the training data may not be learned as completely or correctly, impairing the network's discrimination ability. The degree of impairment is a function of (among other factors) the relative number of samples of each class used for training. The paper addresses the problem of unequal representation in training sets by proposing two alternative methods of learning. One adjusts the learning rate for each class to achieve user-specified goals. The other utilizes a genetic algorithm to set the connection weights with a fitness function based on these same goals. These methods are tested using both artificial and real-world training data.}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {parameters}, applications = {classification} } @article{Dessert92, key = {genetic algorithms connectionism neural networks cogann}, author = {P.E. Dessert}, title = {Anomaly Detection in Data Using Neural Networks With natural selection}, journal = {Proceedings of the SPIE - The International Society for Optical Engineering}, year = {1992}, volume = {1710, pt.1}, pages = {II-725--II-733}, abstract = {ABSTRACT Frequently, time series data taken off machines contains erroneous data points due to errors in the measurement of the data. One such instance of measuring devices recording anomalies occurs in the crash testing of vehicles. Force and acceleration data is collected which an engineer inspects for anomalies, correcting those that are found. Artificial Neural Network (ANN) technology was successfully applied to this problem to eliminate the cost and delay of this manual process. The Author employed " machine learning algorithm that simulates the Darwinian concept of survival of the fittest known as the Genetic Learning Algorithm (GLA). By combining the strength of the GLA and ANNs, a network architecture was created that optimized the size, speed, and accuracy of the ANN. This hybridized system also used the GLA to determine the smallest number of inputs into the ANN that were necessary to detect anomalies in data. This algorithm is known as GENENET, and is described in the paper.}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @incollection{Dodd91, key = {genetic algorithms, connectionism, dolphin vocalization}, author = {N. Dodd}, title = {Optimization of Network Structure using Genetic Techniques}, booktitle = {Applications of Artificial Intelligence in Engineering VI}, year = {1991}, editor = {G. Rzevski and R. A. Adey}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Dodd91a, key = {connectionism}, author = {N. Dodd and D. Macfarlane and C. Marland}, title = {Optimization of Artificial Neural Network Structure Using Genetic Techniques Implemented on Multiple Transputers}, booktitle = {Proceedings of Transputing '91}, year = {1991}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Dolan87, key = {genetic algorithm, connectionism, evolve neural net architecture competitive learning, Hebbian learning, CRAM, cogann ref}, author = {Charles P. Dolan and Michael G. Dyer}, title = {Toward the Evolution of Symbols}, booktitle = {Proceedings of the Second International Conference on Genetic Algorithms}, year = {1987}, editor = {John J. Grefenstette}, publisher = {Lawrence Erlbaum Associates}, address = {Hillsdale, New Jersey}, pages = {123-131}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Dolan87a, key = {genetic algorithm, connectionism, evolve neural net architecture competitive learning, Hebbian learning, CRAM, cogann ref}, author = {Charles P. Dolan and Michael G. Dyer}, title = {Symbolic Schemata in Connectionist Memories: Role Binding and the Evolution of Structure}, institution = {AI Laboratory, University of California, Los Angeles}, year = {1987}, type = {Technical Report UCLA-AI-87-11}, publisher = {UCLA AI Laboratory}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Dominic92, key = {genetic algorithms, connectionism, hill-climbing, mutation only, cogann ref}, author = {S. Dominic and R. Das and D. Whitley and C. Anderson}, title = {Genetic Reinforcement Learning for Neural Networks}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1992}, pages = {II-71 - II-76}, abstract = {Abstract The genetic algorithms which have been shown to yield good performance for neural network weight optimization are really genetic hill-climbers, with a strong reliance on mutation rather than hyperplane sampling. Neural control problems are more appropriate for these genetic hill-climbers than supervised learning applications because in reinforcement learning applications gradient information is not directly available. Genetic reinforcement learning produces competitive results with AHC, another reinforcement learning paradigm for neural networks that employs temporal difference methods. The genetic hill-climbing algorithm appears to be robust over a wide range of learning conditions.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {controller} } @inproceedings{Dress87a, key = {genetic algorithms, connectionism, cogann ref}, author = {W. B. Dress}, title = {Darwinian Optimization of Synthetic Neural Systems}, booktitle = {Proceeding of the IEEE First Annual International Conference on Neural Networks}, year = {1987}, topology = { }, network = { }, encoding = { }, evolves = {connectivity?}, applications = { } } @conference{Dress89, key = {connectionism, cogann ref}, author = {W. B. Dress}, title = {Genetic Optimization in Synthetic Systems}, year = {1989}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @book{Dress90, key = {genetic algorithms, connectionism, cogann ref}, author = {W. B. Dress}, title = {Electronic Life and Synthetic Intelligent Systems}, year = {1990}, publisher = {Instrumentation and Controls Division, Oak Ridge Natuional Laboratory}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Dress90a, key = {genetic algorithms, connectionism, cogann ref}, author = {W. B. Dress}, title = {In-Silico Gene Expression: A Specific Example and Possible Generalizations}, booktitle = {Proceedings of Emergence and Evolution of Life-Forms}, year = {1990}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Dress87, key = {connectionism, genetic algorithms, cogann ref}, author = {W. B. Dress and J. R. Knisley}, title = {A Darwinian Approach to Artificial Neural Systems}, booktitle = {1987 IEEE Conference on Systems, Man, and Cybernetics}, year = {1987}, month = {Oct}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Eberhart91, key = {connectionism, cogann ref}, author = {R. C. Eberhart and R. W. Dobbins}, title = {Designing Neural Network Explanation Facilities Using Genetic Algorithms}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1991}, pages = {1758-1763}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Eberhart92, key = {genetic algorithms, connectionism, neural networks}, author = {Russell C. Eberhart}, title = {The Role of Genetic Algorithms in Neural Network Query-Based Learning and Explanation Facilities}, booktitle = {Proceedings of the Conference on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, pages = {169-183}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @book{Edelman87, author = {Edelman G.M.}, title = {Neural Darwinism: The Theory of Neuronal Group Selection}, year = {1987}, publisher = {Basic Books, New York}, network = { }, applications = { } } @article{Elias92a, key = {genetic algorithms connectionism neural networks cogann}, author = {J.G. Elias}, title = {Target tracking using impulsive analog circuits}, journal = {Proceedings of the SPIE - The International Society for Optical Engineering}, year = {1992}, volume = {1709, pt.1}, pages = {338-350}, abstract = {ABSTRACT The electronic architecture and silicon implementation of an artificial neuron which can be used to process and classify dynamic signals is described. The electrical circuit architecture is modeled after complex neurons in the vertebrate brain which have spatially extensive dendritic tree structures that support large numbers of synapses. The circuit is primarily analog and, as in the biological model system, is virtually immune to process variations and other factors which often plague more conventional circuits. The nonlinear circuit is sensitive to both temporal and spatial signal characteristics but does not make use of the conventional neural network concept of weights, and as such does not use multipliers, adders, look-up-tables, microprocessors or other complex computational devices. The Author shows "ha" artificial neural networks with passive dendritic tree structures can be trained, using a specialized genetic algorithm, to produce control signals useful for target tracking and other dynamic signal processing applications.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {target tracking, signal processing} } @inproceedings{Elias92, key = {genetic algorithms, connectionism, neural networks}, author = {John G. Elias}, title = {Genetic Generation of Connection Patterns for a Dynamic Artificial Neural Network}, booktitle = {Proceedings of the Conference on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, pages = {38-54}, topology = { }, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = { } } @inproceedings{Falcon91, author = {Falcon, J.F.}, title = {Simulated Evolution of Modular Networks}, booktitle = {Artificial Neural Networks, IWANN91, Granada}, year = {1991}, editor = {Prieto, A.}, publisher = {Lecture notes in Computer Science 540, Springer Verlag}, pages = {204-211}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Farmer86, key = {immune networks, machine learning}, author = {Farmer, D. J. and Packard, N. H. and Perelson, A. S.}, title = {The immune system, adaptation, and machine learning}, journal = {Physica}, year = {1986}, volume = {22D}, pages = {187-204}, topology = { }, network = { }, encoding = { }, evolves = {feature detectors}, applications = {pattern classification} } @inproceedings{Fekadu93, author = {Fekadu, A.A. and Hines, E.L. and Gardner, J.W.}, title = {Genetic Algorithm Design of Neural Net Based Electronic Nose}, booktitle = {Artificial Neural Networks and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {691-698}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Feldman93, key = {connectionism cogann}, author = {David S. Feldman}, title = {Fuzzy Network Synthesis and Genetic Algorithms}, booktitle = {Proceedings of the Fifth International Conference on Genetic Algorithms}, year = {1993}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Fenanzo86, key = {genetic algorithms, connectionism, cogann ref}, author = {Fenanzo~Jr, A. J}, title = {Darwinian Evolution as a Paradigm for AI Research}, journal = {SIGART Newsletter}, year = {1986}, month = {July}, number = {97}, pages = {22-23}, publisher = {Harding Lawson Associates}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @conference{Fielder93, key = {genetic algorithms connectionism neural networks cogann}, author = {D. Fielder and C.O. Alford}, title = {Counting and Naming Connection Islands on a Grid of Conductors}, booktitle = {Proceedings of the Conference on Artificial Neural Networks and Genetic Algorithms}, organization = {ANNGA93}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {731}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @mastersthesis{Floreano92, author = {Floreano, D.}, title = {Patterns of Interactions in Ecosystems of Neural Networks}, year = {1992}, school = {Neural Computation, Dept of Comp Sci., Univ of Stirling}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Floreano93, author = {Floreano, D.}, title = {ROBOGEN: a Software Package for Evolutionary Control Systems}, institution = {Cognitive technology laboratory, Trieste}, year = {1993}, number = {93-01}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {control systems robot} } @unpublished{Floreano91, author = {Floreano, D. and Miglino, O. and Parisi, D.}, title = {Emerging Complex Behaviours in Ecosystems of Neural Networks}, year = {1991}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Floreano94, author = {Floreano, D. and Mondada, F.}, title = {Automatic Creation of an Autonomous Agent: Genetic Evolution of a Neural-Network Driven Robot}, booktitle = {Proceedings of the Conference on Simulation of Adaptive Behavior}, year = {1994}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {robot controller} } @article{Fogel90, author = {Fogel, D.B. and Fogel, L.J. and Porto, V.W.}, title = {Evolving Neural Networks}, journal = {Biological Cybernetics}, year = {1990}, volume = {63}, pages = {487-493}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Fogel93a, author = {Fogel, D.B. and Simpson, P.K.}, title = {Evolving Fuzzy Clusters}, booktitle = {Proceedings of the International Conference on Neural Networks}, organization = {ICNN93}, year = {1993}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Fogel93, author = {Fogel, David B.}, title = {Using Evolutionary Programming to Create Neural Networks that are Capable of Playing Tic-Tac-Toe}, booktitle = {Proceedings of the American Power Conference}, year = {1993}, publisher = {IEEE}, pages = {875-879}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {tic-tac-toe} } @inproceedings{Fogel93b, author = {David B. Fogel}, title = {Using Evolutionary Programming to Create Neural Networks that are Capable of Playing Tic-Tac-Toe}, booktitle = {Proceedings of the International Conference on Neural Networks}, year = {1993}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {tic-tac-toe} } @techreport{Fogel93c, key = {genetic algorithms, connectionism, COGANN}, author = {David B. Fogel and Lawrence J. Fogel}, title = {Method and Apparatus for Training a Neural Network Using Evolutionary Programming}, institution = {United States}, year = {1993}, month = {25 MAY}, type = {Patent 5214746}, pages = {731}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @article{Fontanari91, key = {genetic algorithm connectionism}, author = {J.F. Fontanari and R. Meir}, title = {Evolving a Learning Algorithm for the Binary Perceptron}, journal = {Network}, year = {1991}, volume = {2}, pages = {353-359}, network = {perceptron}, encoding = { }, evolves = {learning rule}, applications = { } } @inproceedings{Freisleben93, author = {Freisleben, B. and H\H{a}rtfelder}, title = {Optimization of Genetic Algorithms by Genetic Algorithms}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {392-399}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Fritzke93, author = {Bernd Fritzke}, title = {Growing Cell Structures -- A Self-Organizing Network for Unsupervised and Supervised Learning}, institution = {International Computer Science Institute}, year = {1993}, month = {may}, address = {1947 Center Street, Suit 600, Berkeley, California 94704}, number = {TR-93-026}, topology = {feed-forward}, network = {self-organizing}, encoding = { }, evolves = {feature detectors}, applications = { } } @inproceedings{Fullmer92, key = {genetic algorithms connectionism neural networks cogann}, author = {B. Fullmer and R. Miikkulainen}, title = {Using Marker-Based Genetic Encoding of Neural Networks to Evolve Finite-State Behaviour}, booktitle = {Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life}, year = {1992}, editor = {F.J. Varela and P. Bourgine}, publisher = {MIT Press}, address = {Cambridge, MA, USA}, abstract = {ABSTRACT A new mechanism for genetic encoding of neural networks is proposed, which is loosely based on the marker structure of biological DNA. The mechanism allows all aspects of the network structure, including the number of nodes and their connectivity, to be evolved through genetic algorithms. The effectiveness of the encoding scheme is demonstrated in an object recognition task that requires artificial creatures (whose behavior is driven by a neural network) to develop high-level finite-state exploration and discrimination strategies. The task requires solving the sensory-motor grounding problem, i.e., developing a functional understanding of the effects that a creature's movement has on its sensory input.}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = {object recognition} } @inproceedings{Gallagher92, author = {Gallagher, J. C. and Beer, R. D}, title = {A Qualitative Dynamical Analysis of Evolved Locomotion Control}, booktitle = {From Animals to Animats, Proceedings of the Second International Conference on Simuation of Adaptive Behaviour (SAB 92)}, year = {1992}, editor = {Roitblat, H. and Meyer, J-A. and Wilson, S.}, publisher = {The MIT Press, Cambridge, MA}, topology = {recurrent?}, network = { }, encoding = { }, evolves = { }, applications = {robot controller} } @inproceedings{Game93, author = {Game, G. W. and James, C. D.}, title = {The Application of Genetic Algorithms to the Optimal Selection of Parameter Values in Neural Networks for Attitude Control Systems}, booktitle = {IEE Colloquium on 'High Accuracy Platform Control in Space'}, year = {1993}, publisher = {IEE, London}, pages = {3/1-3/3}, volume = {Digest No. 1993/148}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @techreport{deGaris89, key = {genetic algorithms, connectionism, cogann ref}, author = {Hugo de Garis}, title = {WALKER, A Genetically Programmed, Time Dependent, Neural Net Which Teaches a Pair of Sticks to Walk}, institution = {Center for AI, George Mason Univ, Virginia}, year = {1989}, type = {Technical Report}, publisher = {Center for AI, George Mason Univ, Virginia}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {controller} } @book{deGaris90, key = {connectionism, cogann ref}, author = {Hugo de Garis}, title = {Genetic Programming: Building Nanobrains with Genetically Programmed Neural Network Modules.}, year = {1990}, publisher = {CADEPS AI Research Unit, Universitye Libre de Bruxelles, CP 194/7, B-1050 Brussels, Belgium}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{deGaris90a, key = {genetic algorithm GenNets connectionism, cogann ref}, author = {Hugo de Garis}, title = {BRAIN Building with GenNets}, journal = {Proceedings of INNC-90}, year = {1990}, volume = {2}, pages = {1036-1039}, address = {Paris}, publisher = {Kluwer Academic Publishers}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{deGaris90b, key = {genetic algorithm GenNets connectionism robot control LIZZY, cogann ref}, author = {Hugo de Garis}, title = {Genetic Programming: Evolution of a Time Dependent Neural Network Module which Teaches a Pair of Stick Legs to Walk}, booktitle = {Proceedings of the 9th European Conference on Artificial Intelligence}, year = {1990}, month = {AUG 6-10}, address = {Stockholm, Sweden}, pages = {204-206}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {controller} } @conference{deGaris92, key = {genetic algorithms, connectionism}, author = {Hugo de Garis}, title = {Exploring GenNet Behaviors Using Genetic Programming to Explore Qualitatively New Behaviors in Recurrent Neural Networks}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, organization = {IJCNN-92}, year = {1992}, pages = {III-547 - III-552}, abstract = {ABSTRACT The neural network research community's preoccupation with convergent networks (until the recent rise of 'recurrent backpropagation' algorithms [e.g. WILLIAMS \& ZIPSER 1989ab]) has not been unreasonable. Relatively little analytical work had been done on neural networks whose inputs and/or outputs are time-dependent, hence few guidelines existed on how to train such networks. Consequently, research concentrated on more restrictive 'static' neural nets such as 'feedforward' (Backprop) [RUMELHART \& McCLELLAND 1986] and "Hopfield" (clamped inputs, convergent outputs) [HOPFIELD 1982]. This emphasis on convergence was unfortunate, because the true richness of neural network dynamics is to be found when inputs and/or outputs are time-dependent. This paper shows that Genetic Programming techniques (i.e. using Genetic Algorithms to build/evolve complex systems) can be applied successfully to training nonconvergent networks, and presents some examples of their extraordinary behavioral versatility. This paper terminates by comparing GenNet behaviors with those generated by the new 'recurrent backpropagation' algorithms [WILLIAMS \& ZIPSER 1989ab]. It is claimed that the GenNet behaviors are a lot more flexible and interesting because they do not require the training process to be "closely supervised".}, topology = {recurrent}, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{deGaris92a, key = {genetic algorithms connectionism neural networks cogann}, author = {Hugo de Garis}, title = {Steerable GenNets: the Genetic Programming of Steerable Behaviors in GenNets}, booktitle = {Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life}, year = {1992}, editor = {F.J. Varela and P. Bourgine}, publisher = {MIT Press}, address = {Cambridge, MA, USA}, abstract = {ABSTRACT Shows how genetic programming techniques (i.e. the art of applied evolution, or building complex systems using the genetic algorithm) can be used to evolve dynamic behaviors in neural systems which are controllable or steerable. The genetic algorithm evolves the weights of a fully-connnected time-dependent neural network (called a GenNet), such that the same GenNet is capable of generating two separate time-dependent behaviors, depending upon the setting of two different values of a clamped input control variable. By freezing these weights in the GenNet and then applying intermediate control values, one obtains intermediate behaviors, showing that the GenNet has generalized its behavioral learning. It has become controllable or steerable. This principle is applicable to the evolution of many controllable neural behaviors and is useful in the construction of artificial creatures (with artificial nervous systems) based on neural modules. One simply evolves two behaviors at different settings of the control input so that the GenNet generalizes its behavioral learning. In this paper, a concrete example of this process is given in the form of the genetic programming of a variable frequency generator GenNet. This paper ends with a discussion on the handcrafters vs. evolutionists controversy, concerning future approaches to artificial creature (biot) building.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @conference{deGaris93, key = {genetic algorithms connectionism neural networks cogann}, author = {Hugo de Garis}, title = {Circuits of Production Rule GenNets. The Genetic Programming of Artificial Nervous Systems}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, organization = {ANNGA93}, year = {1993}, pages = {699-705}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @conference{deGaris93a, key = {connectionism, genetic algorithms, cogann, Genetic Programming, GenNets Genetically Programmed Neural Network Modules, Artificial Nervous Systems, Biots Biological Robots, Darwinian Robotics, 1000-GenNet Biots, GenNet Accelerators, GenNet Shaping.}, author = {Hugo de Garis}, title = {Incremental Evolution of Neural Networks: Genetic Programming in Incremental Steps}, booktitle = {Proceedings of the World Congress on Neural Networks}, organization = {WCNN93}, year = {1993}, pages = {II447 - II450}, abstract = {ABSTRACT This paper addresses itself to the question of Incremental Evolution of neural networks, which is defined to be the art of evolving neural networks in incremental steps, using Genetic Algorithms. One evolves the weight values of a fully connected neural network (called a GenNet [de Garis 1990, 1993]) containing N neurons to perform T tasks, and then takes the result (i.e. the evolved weights of the N neurons) and adds a few more neurons dN, to evolve the performance of a few more tasks dT. This paper investigates (a) whether this can be done at all, (b) whether is is faster to evolve an N + dN GenNet performing T + dT tasks from scratch or to do it incrementally (1.e. [N,T] then [N+dN,T+dT]), and (c) how the two approaches (i.e. from scratch or incremental) compare in task performance quality. Incremental Evolution will become an important issue when the various brain builder groups around the world (i.e.groups using evolved neural network modules to build artificial nervous systems for biological robots (biots), e.g. Beer's group at Case Western Reserve University USA, Cliff et al's group at Sussex University UK, and the Author's group "" ATR Japan [de Garis 1993] are confronted with the decision whether to start from scratch when desiring to evolve biots with a greater number of behaviors, or to increment their already evolved nervous systems. Nature obviously had to increment.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Gierer88, key = {genetic algorithms, connectionism, cogann ref}, author = {A. Gierer}, title = {Spatial Organization and Genetic Information in Brain Development}, journal = {Biological Cybernetics}, year = {1988}, volume = {59}, pages = {13-21}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @conference{Gonzalez-Seco92, key = {genetic algorithms, connectionism, GLANN}, author = {Jose Gonzalez-Seco}, title = {A Genetic Algorithm as the Learning Procedure for Neural Networks}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, organization = {IJCNN-92}, year = {1992}, pages = {I-835 - I-840}, abstract = {ABSTRACT The relationship between genetic algorithms and neural networks has been somewhat one directional. In most cases a genetic algorithm has been used to generate better neural networks. In this paper we combine the use of genetics algorithms and neural networks, but from a conceptually different point of view. We show that it is possible to use a genetics algorithm as the learning algorithm for a neural network. In our model the neural network has a fixed architecture and processes binary strings using genetic operators.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Gruau92, author = {Frederic Gruau}, title = {Cellular Encoding of Genetic Neural Network}, institution = {Laboratoire de l'Informatique du Parall\'elisme, Ecole Normale Sup\'erieure de Lyon}, year = {1992}, type = {Research Report 92.21}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity, parameters}, applications = { } } @inproceedings{Gruau92a, author = {Frederic Gruau}, title = {Genetic Synthesis of Boolean Neural Networks with a Cell Rewriting Developmental Process}, booktitle = {Proceedings of COGANN-92 International Workshop on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, editor = {Whitley, L.D. and Schaffer, J.D.}, publisher = {IEEE Computer Society Press}, pages = {55-74}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity, parameters}, applications = { } } @unpublished{Gruau92b, author = {Frederic Gruau}, title = {Cellular Encoding of Genetic Neural Networks I. Theoretical Properties}, year = {1992}, note = {submitted to evolutionnary computation}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Gruau93, author = {Frederic Gruau}, title = {A Learning and Pruning Algorithm for Genetic Neural Networks}, booktitle = {European Symposium on Artificial Neural Networks}, year = {1993}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Gruau93a, key = {genetic algorithms, connectionism cogann}, author = {Frederic Gruau}, title = {Genetic Synthesis of Modular Neural Networks}, booktitle = {Proceedings of the Fifth International Conference on Genetic Algorithms}, year = {1993}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity, parameters}, applications = { } } @incollection{Gruau94, author = {Frederic Gruau}, title = {Genetic Micro Programming of Neural Networks}, booktitle = {Advances in Genetic Programming}, year = {1994}, editor = {Kim Kinnear}, publisher = {MIT Press}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity, parameters}, applications = { } } @phdthesis{Gruau94a, author = {Frederic Gruau}, title = {Neural Network Synthesis Using Cellular Encoding and the Genetic Algorithm}, year = {1994}, school = {PhD Thesis, Ecole Normale Sup\'erieure de Lyon}, note = {anonymous ftp: lip.ens-lyon.fr (140.77.1.11) directory pub/Rapports/PhD file PhD94-01-E.ps.Z (english) PhD94-01-F.ps.Z (french)}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity, parameters}, applications = { } } @techreport{Gruau93b, author = {Frederic Gruau and Darrell Whitley}, title = {The Cellular Development of Neural Networks: the Interaction of Learning and Evolution}, institution = {Laboratoire de l'Informatique du Parallelisme, Ecole Normal Superieure de Lyon}, year = {1993}, number = {93-04}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity, parameters}, applications = { } } @article{Gruau93c, author = {Frederic Gruau and Darrell Whitley}, title = {Adding Learning to the Cellular Developmental Process: a Comparative Study}, journal = {Evolutionary Computation}, year = {1993}, volume = {1}, number = {3}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity}, applications = { } } @techreport{Gruau93d, author = {Frederic Gruau and Darrell Whitley}, title = {Adding Learning to the Cellular Developmental Process: a Comparative Study}, institution = {Laboratoire de l'Informatique du Parall\'elisme, Ecole Normale Sup\'erieure de Lyon}, year = {1993}, type = {Research Report RR93-04}, topology = {general}, network = { }, encoding = {indirect cellular encoding}, evolves = {connectivity}, applications = { } } @article{Gruau93e, key = {genetic algorithm connectionism neural networks cogann}, author = {Fredric Gruau}, title = {Cellular Encoding as a Graph Grammar}, journal = {IEE Colloquium on Grammatical Inference: Theory, Applications and Alternatives}, year = {1993}, month = {22-23 April}, volume = {(Digest No.092)}, pages = {17/1-10}, publisher = {IEE}, address = {London}, abstract = {ABSTRACT Cellular encoding is a method for encoding a family of neural networks into a set of labeled trees. Such sets of trees can be evolved by the genetic algorithm so as to find a particular set of trees that encodes a family of Boolean neural networks for computing a family of Boolean functions. Cellular encoding is presented as a graph grammar. A method is proposed for translating a cellular encoding into a set of graph grammar rewriting rules of the kind used in the Berlin algebraic approach to graph rewriting. The genetic search of neural networks via cellular encoding appears as a grammatical inference process where the language to parse is implicitly specified, instead of explicitly by positive and negative examples. Experimental results shows that the genetic algorithm can infer grammars that derive neural networks for the parity, symmetry and decoder Boolean function of arbitrary large size.}, topology = {general}, network = { }, encoding = {cellular encoding, graph grammar}, evolves = {connectivity, parameters}, applications = {parity etc} } @techreport{Guha92, key = {connectionism, neural networks, cogann}, author = {Aloke Guha and Steven A. Harp and Tariq Samad}, title = {Genetic Algorithm Synthesis of Neural Networks}, institution = {United States}, year = {1992}, month = {18 AUG}, type = {Patent 5140530}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity}, applications = {2-parity, digit recognition, function approximation} } @conference{Guo92, key = {genetic algorithms, connectionism, neural networks}, author = {Zhichao Guo and Robert Uhrig}, title = {Using Genetic Algorithms to Select Inputs for Neural Networks}, booktitle = {Proceedings of the Workshop on Combinations of Genetic Algorithms and Neural Networks}, organization = {COGANN92}, year = {1992}, pages = {223-234}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Hoffgen91, author = {H\H{o}ffgen, K-U. and Siemon, H.P. and Ultsch, A}, title = {Genetic Improvement of Feedforward Nets for Approximating Functions}, booktitle = {Proceedings of the Conference on Parallel Problem Solving from Nature}, year = {1991}, editor = {Schwefel, H-P. and M\H{a}nner, R.}, publisher = {Lecture notes in Computer Science 496, Springer Verlag}, pages = {302-306}, topology = {feed-forward}, network = { }, encoding = { }, evolves = { }, applications = {function approximation} } @inproceedings{Hancock89, author = {Hancock, P. J. B.}, title = {Optimising Parameters in Neural Net Simulations by Genetic Algorithm}, booktitle = {Mini-Symposium on Neural Network Computation}, year = {1989}, publisher = {Rank Prize Funds, Broadway: unpublished}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @inproceedings{Hancock90, key = {connectionism, cogann ref}, author = {P. J. B. Hancock}, title = {GANNET: Design of a Neural Network for Face Recognition by Genetic Algorithm}, booktitle = {Proceedings of the IEEE Workshop on Genetic Algorithms, Simulated Annealing and Neural Networks}, year = {1990}, address = {University of Glasgow, Scotland}, topology = { }, network = { }, encoding = { }, evolves = {connectivity?}, applications = {face recognition} } @phdthesis{Hancock92, author = {Hancock, P. J. B.}, title = {Coding Strategies for Genetic Algorithms and Neural Nets}, year = {1992}, school = {Department of Computing Science and Mathematics, University of Stirling}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Hancock92a, author = {Hancock, P. J. B.}, title = {Recombination Operators for the Design of Neural Nets by Genetic Algorithms}, booktitle = {Parallel Problem Solving from Nature 2}, year = {1992}, editor = {M\H{a}nner, R. and Manderick, B.}, publisher = {Elsevier, North Holland}, pages = {441-450}, topology = { }, network = { }, encoding = { }, evolves = {parameters, connectivity?}, applications = { } } @inproceedings{Hancock92b, author = {Hancock, P. J. B.}, title = {Pruning Neural Nets by Genetic Algorithm}, booktitle = {Proceedings of the International Conference on Artificial Neural Networks, Brighton}, year = {1992}, editor = {Aleksander, I. and Taylor, J.G.}, publisher = {Elsevier}, pages = {991-994}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Hancock92c, author = {Hancock, P. J. B.}, title = {Genetic Algorithms and Permutation Problems: a Comparison of Recombination Operators for Neural Net Structure Specification}, booktitle = {Proceedings of COGANN workshop, IJCNN, Baltimore}, year = {1992}, editor = {Whitley, D.}, publisher = {IEEE}, topology = { }, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = { } } @inproceedings{Hancock91, author = {Hancock, P. J. B. and Smith, L. S}, title = {GANNET: Genetic Design of a Neural Net for Face Recognition}, booktitle = {Proceedings of the Conference on Parallel Problem Solving from Nature}, year = {1991}, editor = {Schwefel, H-P. and M\H{a}nner, R.}, publisher = {Lecture notes in Computer Science 496, Springer Verlag}, pages = {292-296}, topology = { }, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = {face recognition} } @techreport{Harp89, author = {Harp, S. A. and Samad, T. and Guha A.}, title = {The Genetic Synthesis of Neural Networks}, institution = {Honeywell CSDD}, year = {1989}, number = {TR CSDD-89-I4852-2}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity}, applications = {2-parity, digit recognition, function approximation} } @inproceedings{Harp89a, author = {Harp, S. A. and Samad, T. and Guha, A.}, title = {Towards the Genetic Synthesis of Neural Networks}, booktitle = {Proceedings of the Third International Conference on Genetic Algorithms}, year = {1989}, editor = {Schaffer, J.D.}, publisher = {Morgan Kaufmann}, pages = {360-369}, institution = {Honeywell CSDD}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity}, applications = {2-parity, digit recognition, function approximation} } @inproceedings{Harp89b, author = {Harp, S. A. and Samad, T. and Guha, A.}, title = {Designing Application-Specific Neural Networks Using the Genetic Algorithm}, booktitle = {Neural Information Processing Systems 2}, year = {1989}, editor = {Touretzky, D.S.}, institution = {Honeywell CSDD}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity}, applications = {2-parity, digit recognition, function approximation} } @article{Harp92, key = {genetic algorithms connectionism neural networks cogann}, author = {Steven A. Harp and Tariq Samad}, title = {Optimizing Neural Networks with Genetic Algorithms}, journal = {Proceedings of the American Power Conference}, year = {1992}, volume = {54 pt 2}, pages = {1138-1143}, publisher = {Illinois Inst of Technology}, address = {Chicago, IL, USA.}, abstract = {ABSTRACT We describe an approach to application-specific neural network design using genetic algorithms. A genetic algorithm is a robust optimization method particularly well suited for search spaces that are high-dimensional, discontinuous and noisy-features that typify the neural network design problem. Our approach is relevant to virtually all neural network applications: it is network-model independent and it permits optimization for arbitrary, user-defined criteria. We have developed an experimental system, NeuroGENESYS, and have conducted several experiments on small-scale problems. Performance improvements over manual designs have been observed, the interplay between performance criteria and network design aspects has been demonstrated, and general design principles have been uncovered.}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity}, applications = {2-parity, digit recognition, function approximation} } @inproceedings{Harp91a, key = {algorithms, connectionism, Kohonen, clustering, vector quantization, cogann ref}, author = {Steven Alex Harp and Tariq Samad}, title = {Genetic Optimization of Self-Organizing Feature Maps}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1991}, pages = {341-346}, journal = {IJCNN-91}, volume = {I}, network = {self-organizing}, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Harp91, author = {Steven Harp and Tariq Samad}, title = {Genetic Synthesis of Neural Network Architecture}, booktitle = {Handbook of Genetic Algorithms}, year = {1991}, editor = {Davis, L.}, publisher = {Van Nostrand Reinhold}, pages = {202-221}, chapter = {15}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity}, applications = {2-parity, digit recognition, function approximation} } @techreport{Harvey93, author = {Harvey, I. and Husbands, P. and Cliff, D.}, title = {Genetic Convergence in a Species of Evolved Robot Control Architectures}, institution = {Cognitive Science, university of Sussex}, year = {1993}, number = {CSRP 267}, mnote = {See also ICGA93}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {robot controller} } @conference{Hassoun93, key = {connectionism, genetic algorithms, cogann}, author = {Mohamad H. Hassoun and Jing Song}, title = {Multilayer Perceptron Learning Via Genetic Search for Hidden Layer Activations}, booktitle = {Proceedings of the World Congress on Neural Networks}, organization = {WCNN93}, year = {1993}, pages = {III437 - III444}, abstract = {ABSTRACT A new learning technique is proposed for multilayer neural networks based on genetic search, in hidden target space, and gradient descent learning strategies. Our simulations show that the new algorithm combines the global optimization capabilities of genetic algorithms with the speed of gradient descent local search in order to outperform pure descent-based algorithms such as backpropagation. In addition, we show that genetic search in hidden target space is less complex than that of weight space.}, topology = {multi-layered}, network = {multi-layer perceptron}, encoding = { }, evolves = { }, applications = { } } @inproceedings{Heistermann91, author = {Heistermann, J}, title = {The Application of a Genetic Approach as an Algorithm for Neural Networks}, booktitle = {Parallel Problem Solving from Nature}, year = {1991}, editor = {Schwefel, H-P. and M\H{a}nner, R.}, publisher = {Lecture notes in Computer Science 496, Springer Verlag}, pages = {297-301}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Heistermann89, key = {connectionism, genetic algorithms}, author = {J. Heistermann}, title = {Parallel Algorithms for Learning in Neural Networks with Evolution Strategy}, journal = {Parallel Computing}, year = {1989}, volume = {12}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Heistermann90, key = {connectionism}, author = {J. Heistermann}, title = {Learning in Neural Nets by Genetic Algorithms}, booktitle = {Parallel Processing in Neural Systems and Computers}, year = {1990}, editor = {R. Eckmiller and G. Hartmann and G. Hauske}, publisher = {Elsevier Science Publishers}, pages = {165-168}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Heistermann92, key = {genetic algorithms connectionism neural networks cogann}, author = {J. Heistermann}, title = {A Mixed Genetic Approach to the Optimization of Neural Controllers}, booktitle = {CompEuro 1992 Proceedings. Computer Systems and Software Engineering}, year = {1992}, editor = {P. Dewilde and J. Vandewalle}, publisher = {IEEE Comput. Soc. Press}, address = {Los Alamitos, CA, USA}, pages = {459-464}, abstract = {ABSTRACT The Author discusses "om" of the capabilities of genetic algorithms (GAs). GAs are compared with other standard optimization methods like gradient descent or simulated annealing (SA). It is shown that SA is just a special case of GA. The role of a population in the optimization process is demonstrated by an example. GA was applied as a learning algorithm to neural networks.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {controller} } @article{Hinton87, key = {Neural nets genetic algorithms connectionism, cogann ref}, author = {Geoffrey E. Hinton and Stephen J. Nowlan}, title = {How Learning Can Guide Evolution}, journal = {Complex Systems}, year = {1987}, month = {JUN}, volume = {1}, number = {1}, pages = {495-502}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Hintz90, key = {connectionism, genetic algorithm (?)}, author = {K.J. Hintz and J.J. Spofford}, title = {Evolving a Neural Network}, booktitle = {Proceedings of the 5th IEEE International Symposium on Intelligent Control}, year = {1990}, month = {SEPT}, editor = {A. Meystel}, publisher = {IEEE Computer Society Press}, address = {Los Alamitos, CA}, pages = {479-484}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Ho92, key = {genetic algorithms connectionism neural networks cogann}, author = {A.W. Ho and G.C. Fox}, title = {Competitive-Cooperative System of Distributed Artificial Neural Agents}, booktitle = {Parallel Computing: Problems, Methods and Applications. Selection of Papers Presented at the Conference on Parallel Computing: Achievements, Problems and Prospects}, year = {1992}, editor = {P. Messina and A. Murli}, publisher = {Elsevier, Amsterdam, Netherlands}, pages = {499-507}, abstract = {ABSTRACT A framework for simulations of hierarchical organizations of interacting, distributed artificial agents on distributed-memory, MIMD computers is presented. Interactions among aggregates of intelligent agents in an organization are restricted to obey competition and cooperation criteria. Each intelligent agent in an organization is a parallel implementation of a feedforward multilayer perceptrons neural network using error backpropagation (BP) as the learning rule. In this preliminary study, domination, viewed as a type of deterministic genetic algorithm (GA,) is chosen to be the preferred form of interaction. The framework exploits the hierarchical nature intrinsic in an organizational approach to problem-solving. It takes advantage of parallelism at different levels of granularity, from domain decomposition within the agents to coarse grain team-level interaction. Transputer-based simulation results for a test problem of learning the solution to a parity function of predicate order 10 is discussed.}, topology = {multi-layered}, network = {multi-layer perceptron}, encoding = { }, evolves = { }, applications = { } } @article{Holland92, key = {genetic algorithms connectionism neural networks cogann robotics}, author = {O.E. Holland and M.A. Snaith}, title = {Neural Control of Locomotion in a Quadrupedal Robot}, journal = {IEE Proceedings Part F: Radar and Signal Processing}, year = {1992}, month = {DEC}, volume = {139}, number = {6}, pages = {431-436}, abstract = {ABSTRACT The Authors present "esult" of a first study demonstrating that the apparently complex task of controlling walking in a real quadrupedal robot with highly nonlinear interactions between the control elements can be learned quickly by a crude and simple reinforcement learning algorithm. They can as yet say little that is useful about the contribution of reflexes to learned walking, and nothing about the quality of evolved solutions other than that their discovery by applying genetic algorithms to real robots is likely to take a prohibitively long time. However, they hope that their experiences will point the way to more controlled studies of the applications of reinforcement learning to real-world problems, especially to control problems associated with autonomous mobile robots.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {robot controller} } @article{Honavar89a, key = {connectionism neural networks constructive algorithms inductive learning, local architectures, brain modeling, pattern classification}, author = {Honavar, V. and Uhr, L.}, title = {Brain-Structured Networks That Perceive and Learn}, journal = {Connection Science}, year = {1989}, volume = {1}, pages = {139-159}, topology = {feed-forward, locally connected, structured, multi-layered, regular, modular}, network = { }, encoding = { }, evolves = {feature detectors, connectivity, topology}, applications = {pattern classification, vision, brain modeling} } @inproceedings{Honavar89b, key = {connectionism neural networks constructive algorithms inductive learning, pattern classification}, author = {Honavar, V. and Uhr, L.}, title = {A Network of Neuron-Like Units That Learns by Generation As Well As Reweighting of its Links}, booktitle = {Proceedings of the 1988 Connectionist Models Summer School}, year = {1989}, publisher = {Morgan Kaufmann, Palo Alto}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {feature detectors, topology}, applications = {pattern classification} } @inproceedings{Honavar89c, key = {connectionism neural networks constructive algorithms inductive learning, pattern classification}, author = {Honavar, V. and Uhr, L.}, title = {Generation, Local Receptive Fields, and Global Convergence Improve Perceptual Learning in Connectionist Networks}, booktitle = {Proceedings of the Tenth International Joint Conference on Artificial Intelligence}, year = {1989}, publisher = {Morgan Kaufmann, Palo Alto}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {feature detectors, topology}, applications = {pattern classification} } @article{Honavar93, key = {connectionism neural networks constructive algorithms inductive learning, radial basis functions, pattern classification}, author = {Honavar, V. and Uhr, L.}, title = {Generative Learning Structures and Processes for Generalized Connectionist Networks}, journal = {Information Sciences}, year = {1993}, volume = {70}, pages = {75-108}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {feature detectors, connectivity, topology}, applications = {pattern classification} } @inproceedings{Hoptroff90, author = {Hoptroff, R. G. and Hall, T. J. and Burge, R. E.}, title = {Experiments With a Neural Controller}, booktitle = {1990 International Joint Conference on Neural Networks - IJCNN 90}, year = {1990}, publisher = {IEEE, New York}, pages = {735-740}, volume = {2}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {controller} } @inproceedings{Hsu92, author = {Hsu, L.S. and Wu, Z.B.}, title = {Input Pattern Encoding Through Generalized Adaptive Search}, booktitle = {Proceedings of COGANN-92 International Workshop on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, editor = {Whitley, L.D. and Schaffer, J.D.}, publisher = {IEEE Computer Society Press}, pages = {235-247}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Huang92, author = {Huang, R.}, title = {Systems Control With the Genetic Algorithm and the Nearest Neighbour Classification}, journal = {CC-AI}, year = {1992}, volume = {9(2-3)}, pages = {225-236}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {controller} } @techreport{Husbands92, author = {Husbands, P. and Harvey, I. and Cliff, D. T.}, title = {Analysing Recurrent Dynamical Networks Evolved for Robot Control}, institution = {University of Sussex, School of Cognitive and Computing Sciences}, year = {1992}, number = {CSRP265}, topology = {recurrent}, network = { }, encoding = { }, evolves = {connectivity}, applications = {robot controller} } @inproceedings{Ichikawa90, author = {Ichikawa, Y.}, title = {Evolution of Neural Networks and Application to Motion Control}, booktitle = {Proceedings of the IEEE International Conference on Intelligent Motion Control}, year = {1990}, publisher = {IEEE}, pages = {239-245}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = {controller} } @inproceedings{Jacob93, author = {Jacob, C. and Rehder, J.}, title = {Evolution of Neural Net Architectures by a Hierachical Grammar-Based Genetic System}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {72-79}, topology = { }, network = { }, encoding = {indirect, grammar based}, evolves = {connectivity}, applications = { } } @article{Janson92, key = {genetic algorithms connectionism neural networks cogann}, author = {D.J. Janson and J.F. Frenzel}, title = {Application of Genetic Algorithms to the Training of Higher Order Neural Networks}, journal = {Journal of Systems Engineering}, year = {1992}, volume = {2}, number = {4}, pages = {272-276}, abstract = {ABSTRACT Product unit neural networks are a new form of feedforward learning networks in which several summing units are replaced by units capable of calculating a weighted product of inputs. While such networks can be trained using traditional backpropagation, the solution involves the manipulation of complex-valued expressions. As an alternative, this paper investigates the training of product networks using genetic algorithms. Results are presented on the training of a neural network to calculate the optimum width of transistors in a CMOS switch given desired operating parameters. It is shown how local minima affect the performance of the genetic algorithm, and one method of overcoming this is presented.}, topology = {feed-forward}, network = {product-unit networks}, encoding = { }, evolves = {parameters}, applications = { } } @techreport{Jefferson90, author = {Jefferson, D. and Collins, R. and Cooper, C. and Dyer, M. and Flowers, M. and Korf, R. and Taylor, C. and Wang, A.}, title = {Evolution as a Theme in Artificial Life: the Genesys/Tracker System}, institution = {Computer Science, UCLA}, year = {1990}, number = {UCLA-AI-90-09}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Jones93, author = {Jones, A.J.}, title = {Genetic Algorithms and Their Applications to the Design of Neural Networks}, journal = {Neural Computing and Applications}, year = {1993}, volume = {1}, pages = {32-45}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @article{Jones93a, author = {Jones, A.J. and MacFarlane, D.}, title = {Comparing Networks With Differing Neural-Node Functions Using Transputer-Based Genetic Algorithms}, journal = {Neural Computing and Applications}, year = {1993}, volume = {1}, pages = {256-267}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @book{Kadaba90a, key = {vehicle routing, Connectionism, genetic algorithms, XROUTE, expert system, neural network, cogann ref}, author = {Nagesh Kadaba and Kendall E. Nygard}, title = {Improving the Performance of Genetic Algorithms in Automated Discovery of Parameters}, year = {1990}, month = {JAN 25}, publisher = {Dept. of SC and OR, North Dakota State University}, note = {Draft}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @conference{Kargupta91, key = {genetic algorithms selection crowding; relation AI machine learning connectionist networks genetic algorithms, cogann ref}, author = {Hillol Kargupta and Robert E. Smith}, title = {System Identification with Evolving Polynomial Networks}, booktitle = {Proceedings of the Fourth International Conference on Genetic Algorithms}, year = {1991}, pages = {370-376}, abstract = {Abstract: The construction of models for prediction and control of initially unknown, potentially nonlinear systems is a difficult, fundamental problem in machine learning and engineering control. In this paper, a {\em genetic algorithm} (GA) based technique is used to iteratively form polynomial networks that model the behavior of nonlinear systems. This approach is motivated by the {\em group method of data handling} (GMDH) (Ivakhnenko, 1971), but attempts to overcome the computational overhead and locality associated with the original GMDH. The approach presented here uses a multi-modal GA (Deb, 1989) to select nodes for a network based on an information-theoretic fitness measure. Preliminary results show that the GA is successful in modeling continuous-time and discrete-time chaotic systems. Implications and extensions of this work are discussed.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Karim92, key = {process control, connectionism, genetic}, author = {M.N. Karim and S.L. Rivera}, title = {Use of Recurrent Neural Networks for Bioprocess Identification in On-Line Optimization by Micro-Genetic Algorithms}, journal = {Proceedings of the American Control Conference}, year = {1992}, volume = {3}, pages = {1931-1932}, publisher = {American Automatic Control Council}, address = {Green Valley, AZ,}, abstract = {ABSTRACT The use of recurrent neural networks in bioprocess identification and optimization is investigated. A recurrent neural network is trained on a set of fermentation data, and thereafter used as a nonlinear process model to estimate nonmeasurable process states at different conditions. With the bioprocess state variable information available, an optimization technique can be used to generate optimum controls settings to improve the process performance. This paper explores the use of Micro-Genetic Algorithms as a technique for bioreactor optimization. Simulation results will be discussed based in the fermentative ethanol production by the anaerobic bacteria Zymomonas mobilis.}, topology = {recurrent}, network = { }, encoding = { }, evolves = {parameters}, applications = {optimization} } @inproceedings{Karunanithi92, author = {Karunanithi, N. and Das, R. and Whitley, D.}, title = {Genetic Cascade Learning for Neural Networks}, booktitle = {Proceedings of COGANN-92 International Workshop on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, editor = {Whitley, L.D. and Schaffer, J.D.}, publisher = {IEEE Computer Society Press}, pages = {134-145}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @article{Kauffman86, author = {Kauffman, S.A. and Smith, R.G.}, title = {Adaptive Automata Based on Darwinian Selection}, journal = {Physica D}, year = {1986}, volume = {22}, pages = {68-82}, institution = {Univ Pennsylvania}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Keesing91, author = {Keesing, R. and Stork, D.G.}, title = {Evolution and Learning in Neural Networks, the Number and Distribution of Learning Trials Affect the Rate of Evolution}, booktitle = {Advances in Neural Information Processing Systems 3}, year = {1991}, editor = {Lippmann, R.P. and Moody, J.E. and Touretzky, D.S}, publisher = {Morgan Kaufmann}, pages = {804-810}, topology = { }, network = { }, encoding = { }, evolves = {learning rule}, applications = { } } @book{Kerszberg, key = {connectionism, cogann ref}, author = {Michel Kerszberg}, title = {Genetic and Epigenetic Factors in Neural Circuit Wiring (preliminary)}, publisher = {Institut fur Festkorperforschung der}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @incollection{Kerszberg88, key = {genetic algorithms, connectionism, cogann ref}, author = {Michel Kerszberg and Aviv Bergman}, title = {The Evolution of Data Processing Abilities in Competing Automata}, booktitle = {Computer Simulation in Brain Science, Copenhagen, Denmark}, year = {1986}, month = {August}, editor = {Cotterill, R.M.J}, publisher = {Cambridge University Press}, pages = {249-259l}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @phdthesis{Kirby88, key = {genetic algorithms, connectionism, cogann ref}, author = {K.G. Kirby}, title = {Intraneuronal Dynamics and Evolutionary Learning}, year = {1988}, school = {Dept. of Computer Science, Wayne State University}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Kirby86, key = {genetic algorithms, connectionism, cogann ref}, author = {K.G. Kirby and Michael Conrad}, title = {Intraneuronal Dynamics as a Substrate for Evolutionary Learning}, journal = {Physica D}, year = {1986}, volume = {22}, pages = {205-215}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @conference{Kirby89, key = {genetic algorithms, connectionism, cogann ref}, author = {K.G. Kirby and Michael Conrad and R.R. Kampfner}, title = {Evolutionary Learning in Reaction-Diffusion Neurons}, organization = {SUBMITTED TO Bull. Math. Biol.}, year = {1989}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @conference{Kitano90, key = {connectionism, cogann ref}, author = {Hiroaki Kitano}, title = {Empirical Studies on the Speed of Convergence of Neural Network Training Using Genetic Algorithms}, booktitle = {Proceedings of the 8th National Conference on Artificial Intelligence (AAAI-90)}, organization = {PROC AAAI-90}, year = {1990}, publisher = {MIT Press, Cambridge}, pages = {789-795}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @article{Kitano90a, key = {connectionism, cogann ref}, author = {Hiroaki Kitano}, title = {Designing Neural Network Using Genetic Algorithm with Graph Generation System}, journal = {Complex Systems}, year = {1990}, volume = {4}, pages = {461-476}, topology = { }, network = { }, encoding = {graph grammar}, evolves = {connectivity}, applications = { } } @techreport{Kitano92, key = {connectionism, cogann ref}, author = {Hiroaki Kitano}, title = {Neurogenetic Learning: An Integrated Method of Designing and Training Neural Networks using Genetic Algorithms}, institution = {Carnegie Mellon University}, year = {1992}, month = {MAR}, type = {CMU-CMT-92-134}, topology = { }, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = { } } @article{Kitano93, key = {genetic algorithms connectionism neural networks cogann}, author = {Hiroaki Kitano}, title = {Continuous Generation Genetic Algorithms}, journal = {Journal of the Society of Instrument and Control Engineers}, year = {1993}, month = {Jan.}, volume = {32}, number = {1}, pages = {31-8}, abstract = {ABSTRACT Presents a continuous generation genetic algorithm. Most genetic algorithms use a discrete generation model in which all individuals in a population synchronize mating period. The discrete generation model, however, wastes processor time in parallel implementations when the fitness of each individual (proportionally or reversely) correlates with the computational cost of its evaluation. An example of such a task is neural network design and training. In some cases, over 80been wasted. The continuous generation model mitigates this problem by introducing asynchronous mating, the continuous generation model increases the number of reproduction per a unit-time over 500discrete model. CPU idle time has been minimized to 1/25. Also, a significant improvement in convergence speed has been estimated.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Kouchi92, key = {genetic algorithms connectionism neural networks cogann}, author = {M. Kouchi and H. Inayoshi and T. Hoshino}, title = {Optimization of Neural-Net Structure by Genetic Algorithm with Diploidy and Geographical Isolation Model}, journal = {Journal of Japanese Society for Artificial Intelligence}, year = {1992}, month = {May}, volume = {7}, number = {3}, pages = {509-517}, abstract = {ABSTRACT The structure of a simple neural network is optimized by the use of a genetic-algorithm. The neural network is a perceptron, which has three outputs; the logical AND, OR and XOR of two inputs The evaluation function for optimization is a linear combination of the correctness, the network sizes, and an auxiliary term inducing the optimum solution The chromosome is a vector of the link weights of the network. The genetic operators used are crossing-over and point-mutation on the parent chromosomes Two genetic rules were tested. In the haploidy rule, each individual has single chromosome, and the offspring is generated by crossing-over the parents' chromosomes at a randomly chosen locus and taking one of those crossed-over chromosomes. In the diploidy rule, each individual has a pair of chromosomes, and the offspring's chromosomes are generated by combining the gamete produced through the meiosis of the parents' chromosomes. The other model used in the genetic algorithm is the geographical isolation model, where the entire population is divided into four sub-populations, in which the local selection and reproduction are carried out, though, in some time interval, randomly sampled individuals are exchanged among sub-populations. Comparison was made among four combinations of haploid or diploid, and single-population or multiple sub-populations. Diploidy together with the sub-population model was proved to be the best for this optimization problem. Thus, the optimum structure of network was found.}, network = {perceptron}, encoding = { }, evolves = {connectivity}, applications = { } } @book{Koza92, author = {John R. Koza}, title = {Genetic Programming: On the Programming of Computers by Means of Natural Selection}, year = {1992}, publisher = {MIT Press, Cambridge, Mass.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Koza91, key = {genetic algorithms, connectionism, one-bit adder, cogann ref}, author = {John R. Koza and James P. Rice}, title = {Genetic Generation of Both the Weights and Architecture for a Neural Network}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, organization = {IJCNN-91}, year = {1991}, pages = {397-404}, journal = {IJCNN-91}, volume = {II}, topology = { }, network = { }, encoding = {genetic programming?}, evolves = {connectivity, parameters}, applications = { } } @inproceedings{Krishnakumar92, author = {Krishnakumar, K.}, title = {Immunized Neurocontrol - Concepts and Initial Results}, booktitle = {Proceedings of COGANN-92 International Workshop on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, editor = {Whitley, L.D. and Schaffer, J.D.}, publisher = {IEEE Computer Society Press}, pages = {146-168}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Kwiatkowski93a, author = {Kwiatkowski, L. and Stromboni, J.P.}, title = {Neuromimetic Algorithm processing: Tools for Design of Dedicated Architectures}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {706-711}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Lai92, key = {genetic algorithms, connectionism}, author = {W.K. Lai and G.G. Coghill}, title = {Genetic Breeding of Control Parameters for the Hopfield/Tank Neural Net}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, organization = {IJCNN-92}, year = {1992}, pages = {IV-618 - IV-623}, abstract = {ABSTRACT Artificial neural networks, especially the Hopfield/Tank neural net have been used to solve the travelling salesman problem. These networks usually require a set of parameters to be carefully selected and tuned to produce sensible solutions. Genetic Algorithms are basically adaptive systems that transform a population of individuals into new populations, using relatively simple mechanisms. It has the ability to efficiently explore the problem sub-space to produce approximate solutions that are globally competitive. This paper will show how Genetic Algorithms may be used in conjunction with the Hopfield/Tank neural net by breeding an effective set of control parameters in the parameter sub-space to be used by the artificial neural network.}, topology = {hopfield network}, network = { }, encoding = { }, evolves = {parameters}, applications = {travelling salesperson problem} } @mastersthesis{Lange93, author = {Frank Lange}, title = {"Uber den Zusammenhang zwischen Komplexit"at und Generalisierungsf"ahigkeit Neuronaler Netze}, year = {1993}, school = {Universit"at Karlsruhe, Institut f"ur Logik, Komplexit"at und Deduktionssysteme}, note = {Beyond Soft-Weight-Sharing: Soft-Entropy-Minimization}, type = {Diplomarbeit}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Lehar87, author = {Lehar, S. and Weaver, J.}, title = {A Developmental Approach to Neural Network Design}, booktitle = {Proceedings of the IEEE International Conference on Neural Networks}, year = {1987}, publisher = {IEEE Press}, pages = {97-104}, topology = { }, network = { }, encoding = { }, applications = { } } @inproceedings{Lewis92a, key = {genetic algorithms connectionism neural networks cogann programming}, author = {M. Anthony Lewis and Andrew H. Fagg and Alan Solidum}, title = {Genetic Programming Approach to the Construction of a Neural Network for Control of a Walking Robot}, booktitle = {Proceedings of IEEE International Conference on Robotics and Automation}, year = {1992}, publisher = {IEEE}, address = {Piscataway, NJ, USA}, pages = {2618-2623}, volume = {3}, abstract = {ABSTRACT The Authors describe "h" staged evolution of a complex motor pattern generator (MPG) for the control of a walking robot. The experiments were carried out on a six-legged, Brooks-style insect robot. The MPG was composed of a network of neurons with weights determined by genetic algorithm optimization. Staged evolution was used to improve the convergence rate of the algorithm. First, an oscillator for the individual leg movements was evolved. Then, a network of these oscillators was evolved to coordinate the movements of the different legs. By introducing a staged set of manageable challenges, the algorithm's performance was improved.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {robot controller} } @inproceedings{Lindgren92a, author = {Lindgren, K. and Nilsson, A. and Nordahl, M.G. and Rade, I.}, title = {Regular Language Inference Using Evolving Neural Networks}, booktitle = {Proceedings of COGANN-92 International Workshop on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, editor = {Whitley, L.D. and Schaffer, J.D.}, publisher = {IEEE Computer Society Press}, pages = {75-86}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {regular language inference} } @inproceedings{Lindgren93a, author = {Lindgren, K. and Nilsson, A. and Nordahl, M.G. and Rade, I.}, title = {Evolving Recurrent Neural Networks}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {55-62}, topology = {recurrent}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Littman91, key = {genetic algorithms environment fitness functions dynamic; biological modeling evolution and learning; hillclimbing, cogann ref ERL, Evolutionary reinforcement, non-stationary environment, dynamic , neural networks, connectionism}, author = {Michael L. Littman and David H. Ackley}, title = {Adaptation in Constant Utility Non-Stationary Environments}, booktitle = {Proceedings of the Fourth International Conference on Genetic Algorithms}, year = {1991}, pages = {136-142}, abstract = {Abstract: Environments that vary over time present special challenges to adaptive systems. Although in the worst case there may be no hope of effective adaptation, not all forms of environmental variability need be so disabling. We consider a broad class of non-stationary environments, those which combine a variable *result function* with an invariant *utility function*, and demonstrate via simulation that an adaptive strategy employing both evolution and learning can tolerate a much higher rate of environmental variation than an evolution-only strategy. We suggest that in many cases where stability has previously been assumed, the constant utility non-stationary environment may in fact be a more robust description.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Muhlenbein90a, author = {Heinz M\H{u}hlenbein}, title = {Limitations of Multi-Layer Perceptron Networks - Steps Towards Genetic Neural Networks}, journal = {Parallel Computing}, year = {1990}, volume = {14}, pages = {249-260}, topology = {multi-layered}, network = {multi-layer perceptron}, encoding = { }, evolves = { }, applications = { } } @inproceedings{Muhlenbein92, key = {genetic algorithms connectionism neural networks cogann}, author = {Heinz M\H{u}hlenbein}, title = {Parallel Genetic Algorithms and Neural Networks as Learning Machines}, booktitle = {Parallel computing '91 Proceedings of the International Conference}, year = {1992}, editor = {D. J. Evans and G. R. Joubert and H. Liddell}, publisher = {North-Holland Publishing Co.}, address = {Amsterdam}, pages = {91-103}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @incollection{Muhlenbein89a, key = {connectionism, cogann ref}, author = {Heinz M\H{u}hlenbein and J{\"o}rg Kindermann}, title = {The Dynamics of Evolution and Learning: Towards Genetic Neural Networks}, booktitle = {Connectionism in Perspective}, year = {1989}, editor = {R. Pfeifer and Z. Schreter and F. Fogelman-Soulie and L. Steels}, publisher = {Elsevier Science Publishers B.V. (North-Holland)}, pages = {173-197}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @book{Machado92, key = {genetic algorithms connectionism neural networks cogann}, author = {Ricardo Jose Machado and Armando Freitas da Rocha}, title = {Evolutive Fuzzy Neural Networks}, year = {1992}, publisher = {IEEE}, address = {Piscataway, NJ}, journal = {1992 IEEE INT CONF Fuzzy Syst FUZZ-IEEE}, pages = {493-500}, abstract = {ABSTRACT The Authors describe "h" combination of fuzzy neural networks with genetic algorithms, producing a flexible and powerful learning paradigm, called evolutive learning. Evolutive learning combines as complementary tools both inductive learning through synaptic weight adjustment and deductive learning through the modification of the network topology to obtain the automatic adaptation of system knowledge to the problem domain environment. Algorithms for the development of an evolutive learning machine are presented. A fuzzy criterion based on entropy is proposed to select the architecture for a fuzzy neural network best suited to a specific problem domain.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Maeda92, key = {genetic algorithms connectionism neural networks cogann}, author = {Y. Maeda and Y. Kanata}, title = {A Genetic Algorithm for an Unsupervised Learning of Neural Networks}, journal = {Engineering \& Technology}, year = {1992}, volume = {10}, number = {2}, pages = {1-7}, abstract = {ABSTRACT The Authors deal "it" a genetic algorithm for an unsupervised learning rule of neural networks. The genetic algorithm consists of four operations: selection; reproduction; crossover; and mutation. They look into the learning efficiency of two kinds of the crossover for the unsupervised learning rule. Moreover, they investigate the learning rate with respect to the mutation rate.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @incollection{Mandischer93a, author = {M.~Mandischer}, title = {Representation and Evolution of Neural Networks}, booktitle = {Artificial Neural Nets and Genetic Algorithms Proceedings of the International Conference at Innsbruck, Austria}, year = {1993}, editor = {R.F.~Albrecht and C.R.~Reeves and N.C.~Steele}, publisher = {Springer}, address = {Wien and New York}, pages = {643--649}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity}, applications = { } } @inproceedings{Maniezzo93, author = {Maniezzo, V.}, title = {Searching Among Search Spaces: Hastening the Genetic Evolution of Feedforward Neural Networks.}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {635-643}, topology = {feed-forward}, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Maniezzo94a, author = {Maniezzo, V.}, title = {Genetic Evolution of the Topology and Weight Distribution of Neural Networks}, journal = {IEEE Transactions on Neural Networks}, year = {1994}, volume = {5}, pages = {39-53}, topology = { }, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = { } } @inproceedings{Maricic90, key = {connectionism}, author = {Borut Maricic and Zoran Nikolov}, title = {GENNET - System for Computer Aided Neural Network Design Using Genetic Algorithms}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1990}, month = {Jan}, address = {Washington, DC}, pages = {I-102 - I-105}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @article{Marin93, author = {F.J. Marin and F. Sandoval}, title = {Genetic Synthesis of Discrete-Time Recurrent Neural Network}, journal = {New Trends in Neural Computation, Springer-Verlag}, year = {1993}, pages = {179-184}, topology = {recurrent}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Marti92, key = {genetic algorithms, connectionism}, author = {Leonardo Mart{\'i}}, title = {Genetically Generated Neural Networks II: Searching for an Optimal Representation}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1992}, pages = {II-221 - II-226}, abstract = {ABSTRACT Genetic Algorithms (GAs) make use of an internal representation of a given system in order to perform optimization functions. The actual structural layout of this representation, called a genome, has a crucial impact on the outcome of the optimization process. The purpose of this paper is to study the effects of different internal representations in a GA, which generates neural networks. A second GA was used to optimize the genome structure. This structure produces an optimized system within a shorter time interval.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Marti92a, key = {genetic algorithms, connectionism}, author = {Leonardo Mart{\'i}}, title = {Genetically Generated Neural Networks I: Representational Effects}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, organization = {IJCNN-92}, year = {1992}, pages = {IV-537 - IV-542}, abstract = {ABSTRACT This paper studies several applications of genetic algorithms (GAs) within the neural networks field. After generating a robust GA engine, the system was used to generate neural network circuit architectures. This was accomplished by using the GA to determine the weights in a fully interconnected network. The importance of the internal genetic representation was shown by testing different approaches. The effects in speed of optimization of varying the constraints imposed upon the desired network were also studied. It was observed that relatively loose constraints provided results comparable to a fully constrained system. The typeof neural network circuits generated were recurrent competitive fields as described by Grossberg (1982).}, topology = {recurrent}, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @mastersthesis{Mayer93, key = {genetic algorithms connectionism neural networks cogann}, author = {Erik Mayer}, title = {Genetic Algorithm Approach to Neural Network Optimization}, year = {1993}, month = {August}, address = {Toledo, Ohio}, school = {University of Toledo}, type = {Masters Thesis}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @article{Maynard87, author = {Maynard Smith, J.}, title = {When Learning Guides Evolution}, journal = {Nature}, year = {1987}, volume = {329}, pages = {761-762}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{McDonnell92, key = {genetic algorithms connectionism neural networks cogann evolutionary programming}, author = {John R. McDonnell and Don E. Waagen}, title = {Evolving Neural Network Architecture}, journal = {Proceedings of SPIE - The International Society for Optical Engineering}, year = {1992}, volume = {1766}, pages = {690-701}, publisher = {Society for Optical Engineering}, address = {Bellingham, WA USA}, abstract = {ABSTRACT This work investigates the application of a stochastic search technique, evolutionary programming, for developing self-organizing neural networks. The chosen stochastic search method is capable of simultaneously evolving both network architecture and weights. The number of synapses and neurons are incorporated into an objective function so that network parameter optimization is done with respect to computational costs as well as mean pattern error. Experiments are conducted using feedforward networks for simple binary mapping problems.}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = {binary mapping} } @conference{McDonnell93, key = {connectionism, genetic algorithms, cogann}, author = {John R. McDonnell and Don E. Waagen}, title = {Determining Neural Network Hidden Layer Size Using Evolutionary Programming}, booktitle = {Proceedings of the World Congress on Neural Networks 93}, organization = {WCNN93}, year = {1993}, pages = {III564 - III567}, abstract = {ABSTRACT This work investigates the application of evolutionary programming, a stochastic search technique, for simultaneously determining the weights and the number of hidden units in a fully-connected, multi-layer neural network. The simulated evolution search paradigm provides a means for optimizing both network structure and weight coefficients. Orthogonal learning is implemented by independently modifying network structure and weight parameters. Different structural level search strategies are investigated by comparing the training processes for the 3-bit parity problem. The results indicate that evolutionary programming provides a robust framework for evolving neural networks.}, topology = {general}, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = {3-parity} } @inproceedings{McDonnell93a, author = {McDonnell, John R. and Don E. Waagen}, title = {Evolving Neural Network Connectivity}, booktitle = {Proceedings of the American Power Conference}, year = {1993}, publisher = {IEEE}, pages = {863-868}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @article{McDonnell94, author = {John R. McDonnell and Don E. Waagen}, title = {Evolving Recurrent Perceptrons for Time-Series Modeling}, journal = {IEEE Transactions on Neural Networks}, year = {1994}, volume = {5}, pages = {24-38}, topology = {recurrent perceptron}, network = { }, encoding = { }, evolves = { }, applications = {time-series} } @inproceedings{McGregor92, author = {McGregor, D. R. and Odetayo, M. O. and Dasgupta, D.}, title = {Adaptive-Control of a Dynamic System Using Genetic-Based Methods}, booktitle = {Proceedings of the 1992 IEEE International Symposium on Intelligent Control}, year = {1992}, publisher = {IEEE}, pages = {521-525}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {control system} } @conference{McInerney, key = {connectionism cogann xor encoder}, author = {Michael McInerney and Atam P. Dhawan}, title = {Use of Genetic Algorithms with Back Propagation in Training of Feed-Forward Neural Networks}, organization = {preprint from the Author; don't "no" about publication}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {parameters}, applications = {2-parity, encoder-decoder} } @incollection{Menczer94, author = {Menczer, F. and Belew, R.K.}, title = {Latent Energy Environments}, booktitle = {Plastic Individuals in Evolving Populations}, year = {1994}, editor = {Belew, R.K. and Mitchell, M.}, publisher = {Addison Wesley}, address = {Reading, MA}, note = {(in press)}, series = {Santa Fe Institute Studies in the Sciences of Complexity}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Menczer94a, author = {Menczer, F. and Belew, R.K.}, title = {Evolving Sensors in Environments of Controlled Complexity}, booktitle = {Artificial Life IV}, year = {1994}, editor = {Brooks, R. and Maes, P.}, publisher = {MIT Press}, address = {Cambridge, MA}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Menczer90, author = {Menczer, F. and Parisi, D.}, title = {`Sexual' Reproduction in Neural Networks}, institution = {C.N.R.Rome}, year = {1990}, number = {PCIA-90-06}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Menczer92, key = {genetic algorithms connectionism neural networks cogann}, author = {F. Menczer and D. Parisi}, title = {Recombination and Unsupervised Learning: Effects of Crossover in the Genetic Optimization of Neural Networks}, journal = {Network: Computation in Neural Systems}, year = {1992}, month = {Nov}, volume = {3}, number = {4}, pages = {423-442}, abstract = {ABSTRACT Genetic algorithms have been successfully used for optimizing complex functions over multidimensional domains, such as the space of the connection weights in a neural network. A feed-forward layered network is used to simulate the life cycle of a synthetic animal that moves in an environment and captures food objects. The adaptation of the animal (i.e. of the network's weight matrix) to the environment can be measured by the amount of reached food objects in a given lifetime. The Authors consider "hi" amount as a fitness function to be optimized by a genetic algorithm over the space of the connection weights. The network can learn the weights that solve the survival task only by means of its genetic evolution. The recombination genetic operator (crossover) can be seen as a model of sexual recombination for the population, while mutation models agamic reproduction. The central problem in trying to apply crossover is the difficult mapping between the genetic code string (genotype) and the network's weight matrix (phenotype). For this reason crossover has been considered unsuitable for this kind of problem in the past. The Authors propose " simple mapping and compare the effects of sexual versus agamic reproduction in such a problem. The results of several parametric simulations are outlined, showing that crossover actually helps to speed up the genetic learning.}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {parameters}, applications = {simulated world} } @inproceedings{Menczer92a, key = {genetic algorithms connectionism neural networks cogann}, author = {F. Menczer and D. Parisi}, title = {A Model for the Emergence of Sex in Evolving Networks: Adaptive Advantage or Random Drift?}, booktitle = {Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life}, year = {1992}, editor = {F.J. Varela and P. Bourgine}, publisher = {MIT Press}, address = {Cambridge, MA, USA}, pages = {337-345}, abstract = {ABSTRACT The evolution of sex is an intriguing problem in evolutionary biology: most high organisms use some form of sexual recombination of the genetic material in the process of reproduction, thus there should be an adaptive advantage in recombination if sex was selected in the course of evolution. One might hope that the new tools offered by the simulation methods of artificial life, genetic algorithms, (GA) and neural networks, might help the investigation by allowing the study of simplified models and of their detailed consequences. The Authors start "ro" some results on the effects of introducing crossover in a GA used for evolving a population of artificial animals trained on a simple task. Since there is a clear advantage in applying crossover versus simple mutations alone, this advantage could be retained by the population through selection: this hypothesis is tested in a model with local, individual genetic operators' probabilities by studying the emergent recombination frequencies. It is unexpectedly hard to analyze the results of the simulations, as the operator probabilities do not enter directly in the computation of fitness, while they have a well-known indirect influence on the 'behaviour' of fitness. The Authors are "onitorin" a trait that is not directly selected, thus being subject to the strong action of random drift.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Menczer92b, key = {genetic algorithms connectionism neural networks cogann}, author = {F. Menczer and D. Parisi}, title = {Evidence of Hyperplanes in the Genetic Learning of Neural Networks}, journal = {Biological Cybernetics}, year = {1992}, volume = {66}, number = {3}, pages = {283-289}, abstract = {ABSTRACT Genetic algorithms (GA) have been successfully applied to the learning process of neural networks simulating artificial life. In previous research the Authors (1990) "ompare" mutation and crossover as genetic operators on neural networks directly encoded as real vectors. With reference to crossover they were actually testing the building blocks hypothesis, as the effectiveness of recombination relies on the validity of such hypothesis. Even with the real genotype used, it was found that the average fitness of the population of neural networks is optimized much more quickly by crossover than it is by mutation. This indicated that the intrinsic parallelism of crossover is not reduced by the high cardinality. In this paper the Authors first "ummariz" such findings and then propose an interpretation in terms of the spatial correlation of the fitness function with respect to the metric defined by the average steps of the genetic operators. Some numerical evidence of such interpretation is given, showing that the fitness surface appears smoother to crossover than it does to mutation. This confirms indirectly that crossover moves along privileged directions, and at the same time provides a geometric rationale for hyperplanes.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Merelo93, author = {Merelo and Paton and Canias and Prieto and Moran}, title = {Genetic Optimization of a Multilayer Neural Network for Cluster Classification Tasks}, journal = {Neural Network World}, year = {1993}, pages = {175-186}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {pattern classification} } @inproceedings{Miglino93a, author = {Miglino, O. and Pedone, R. and Parisi, D.}, title = {A 'Noise Gene' for Econets}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {588-594}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Mikami93, author = {Mikami, S. and Tano, H. and Kakazu, Y.}, title = {An Autonomous Legged Robot That Learns to Walk Through Simulated Evolution}, booktitle = {Self-Organisation and Life, From Simple Rules to Global Complexity, Proceedings of the Second European Conference on Artificial Life}, year = {1993}, publisher = {MIT Press, Cambridge}, pages = {758-767}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {robot controller} } @inproceedings{Miller89, author = {Miller, G.F. and Todd, P.M. and Hegde, S.U.}, title = {Designing Neural Networks Using Genetic Algorithms}, booktitle = {Proceedings of the Third International Conference on Genetic Algorithms}, year = {1989}, editor = {Schaffer, J.D.}, publisher = {Morgan Kaufmann}, pages = {379-384}, topology = {feed-forward}, network = { }, encoding = {direct}, evolves = {connectivity}, applications = {parity, four-quadrant, pattern copying} } @article{Mitchell93, author = {Mitchell, M. and Forrest, S.}, title = {Genetic Algorithms and Artificial Life}, journal = {Artificial Life}, year = {1993}, mnote = {Santa Fe Institute Working Paper 93-11-072 Genetic algorithms are computational models of evolution that play a central role in many artificial-life models. We review the history and current scope of research on genetic algorithms in artificial life, using illustrative examples in which the genetic algorithm is used to study how learning and evolution interact, and to model ecosystems, immune system, cognitive systems, and social systems. We also outline a number of open questions and future directions for genetic algorithms in artificial-life research.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Mitchell93a, author = {Mitchell, R.J. and Bishop, J.M. and Low, W.}, title = {Using a Genetic Algorithm to Find the Rules of a Neural Network}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {664-669}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Mjolsness86, key = {connectionism, genetic algorithms, cogann ref}, author = {E. Mjolsness and D. H. Sharp}, title = {A Preliminary Analysis of Recursively Generated Networks}, booktitle = {Proceedings of the American Institute of Physics (Special Issue on Neural Nets)}, year = {1986}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Mjolsness88, key = {connectionism, genetic algorithms, cogann ref}, author = {Eric Mjolsness and David H. Sharp and Bradley K. Alpert}, title = {Scaling, Machine Learning, and Genetic Neural Nets}, year = {1988}, month = {March}, address = {Theoretical Division, Los Alamos National Laboratory}, type = {LA-UR-88-142}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inbook{Mjolsness88a, key = {genetic connectionism, cogann ref}, author = {Eric Mjolsness and David H. Sharp and Bradley K. Alpert}, title = {Genetic Parsimony in Neural Nets}, year = {1988}, booktitle = {Snowbird 1988}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Mjolsness89, key = {connectionism, cogann ref}, author = {Eric Mjolsness and David H. Sharp and Bradley K. Alpert}, title = {Scaling, Machine Learning, and Genetic Neural Nets}, journal = {Advances in Applied Mathematics}, year = {1989}, volume = {10}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Montana89, author = {Montana, D.J. and Davis, L.}, title = {Training Feedforward Neural Networks Using Genetic Algorithms}, booktitle = {Proceedings of the Eleventh International Joint Conference on Artificial Intelligence}, year = {1989}, publisher = {San Mateo, CA: Morgan Kaufmann.}, pages = {762-767}, institution = {BBN Systems}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @techreport{Moriarty93, author = {David E. Moriarty and Risto Miikkulainen}, title = {Evolving Complex {O}thello Strategies Using Marker-based Genetic Encoding of Neural Networks}, institution = {Department of Computer Sciences, The University of Texas at Austin}, year = {1993}, number = {AI93-206}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {game playing (othello)} } @techreport{Moriarty94, author = {David E. Moriarty and Risto Miikkulainen}, title = {Using Evolutionary Neural Networks for Value Ordering in Constraint Satisfaction Problems}, institution = {Deptartment of Computer Sciences, The University of Texas at Austin}, year = {1994}, number = {AI94-218}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Moriarty94a, author = {David E. Moriarty and Risto Miikkulainen}, title = {Evolving Neural Networks to Focus Minimax Search}, booktitle = {Proceedings of the Twelfth National Conference on Artificial Intelligence (AAAI-94)}, year = {1994}, address = {Seattle, WA}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Munro93, author = {Munro, P.}, title = {Genetic Search for Optimal Representations in Neural Networks}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {628-634}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Nagao92, key = {genetic algorithms connectionism neural networks cogann}, author = {T. Nagao and T. Agui and H. Nagahashi}, title = {Structural Evolution of Neural Networks by a Genetic Method}, journal = {Transactions of the Institute of Electronics, Information and Communication Engineers D-II}, year = {1992}, month = {Sep}, volume = {J75D-II}, number = {9}, pages = {1634-1637}, abstract = {ABSTRACT A method of neural networks construction by a genetic algorithm is proposed. Each network has mutual connections and is assumed to be a living thing whose genes denote the connections among its units. In order to find out a network available to the current task, the simulation of evolution processes of the networks is executed.}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @article{Nagao93, author = {Nagao, T. and Agui, T. and Nagahashi, H.}, title = {Structural Evolution of Neural Networks Having Arbitrary Connection by a Genetic Method}, year = {1993}, volume = {E76-D(6)}, pages = {689-697}, booktitle = {IEICE Transactions on Information and Systems}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Nagao93a, author = {Nagao, T. and Agui, T. and Nagahashi, H.}, title = {A Genetic Method for Optimization of Asynchronous Random Neural Networks and its Application to Action Control}, booktitle = {IJCNN'93-NAGOYA Proceedings of the 1993 International Joint Conference on Neural Networks, Nagoya (Japan)}, year = {1993}, publisher = {IEEE}, pages = {2869-2872}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Narayanan93, key = {genetic algorithms connectionism neural networks cogann}, author = {M.N. Narayanan and S.B. Lucas}, title = {A Genetic Algorithm to Improve a Neural Network to Predict a Patient's Response to Warfarin}, journal = {Methods of Information in Medicine}, year = {1993}, volume = {32}, number = {1}, pages = {55-8}, abstract = {ABSTRACT The ability of neural networks to predict the international normalised ratio (INR) for patients treated with Warfarin was investigated. Neural networks were obtained by using all the predictor variables in the neural network, or by using a genetic algorithm to select an optimal subset of predictor variables in a neural network. The use of a genetic algorithm gave a marked and significant improvement in the prediction of the INR in two of the three cases investigated. The mean error in these cases, typically, reduced from 1.02+or-0.29 to 0.28+or-0.25 (paired t-test, t=-4.71, p<0.001, n=30). The use of a genetic algorithm with Warfarin data offers a significant enhancement of the predictive ability of a neural network with Warfarin data, identifies significant predictor variables, reduces the size of the neural network and thus the speed at which the reduced network can be trained, and reduces the sensitivity of a network to over-training.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Nolfi90, author = {Nolfi, S. and Elman, J.L. and Parisi, D.}, title = {Learning and Evolution in Neural Networks}, institution = {UCSD}, year = {1990}, month = {July}, number = {CRL TR 9019}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Nolfi91, author = {Nolfi, S. and Parisi, D.}, title = {Auto-Teaching Networks That Develop Their Own Teaching Input}, institution = {Dept. of Cognitive Processes and Artificial Intelligence}, year = {1991}, number = {PCIA-91-03}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Nolfi91a, author = {Nolfi, S. and Parisi, D.}, title = {Growing Neural Networks}, institution = {Institute of Psychology, CNR, Rome}, year = {1991}, number = {PCIA-91-15}, note = {Also in Proceedings of ALIFE III, 1992}, topology = {feed-forward}, network = { }, encoding = {indirect, developmental}, evolves = {connectivity, parameters}, applications = {simulated world} } @inproceedings{Nolfi93, author = {Nolfi, S. and Parisi, D.}, title = {Self-Selection of Input Stimuli for Improving Performance}, booktitle = {Neural Networks In Robotics}, year = {1993}, editor = {Bekey, G.A. and Goldberg, K.Y.}, publisher = {Kluwer academic publishers}, pages = {403-420}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Neill92, key = {genetic algorithms connectionism neural networks cogann}, author = {A.W. O'Neill}, title = {Genetic Based Training of Two-Layer, Optoelectronic Neural Network}, journal = {Electronics Letters}, year = {1992}, month = {Jan}, volume = {28}, number = {1}, pages = {47-48}, abstract = {ABSTRACT For the first time, the supervised training of a high-speed, two-layer, optoelectronic neural network using a genetic algorithm is demonstrated, and results for the 3 bit exclusive-or function are presented.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = {3-parity} } @incollection{Ohsuga91, author = {Setsuo Ohsuga and Hannu Kangassalo and Hannu Jaakkola and Koichi Hori and N. Yonezaki}, title = {Information Modeling and Knowledge Bases: Foundations, Theory, and Applications}, year = {1991}, editor = {Setsuo Ohsuga and Hannu Kangassalo and Hannu Jaakkola and Koichi Hori and N. Yonezaki}, publisher = {{IOS} Press}, address = {Amsterdam}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Oliker92, key = {genetic algorithms connectionism neural networks cogann}, author = {S. Oliker and M. Furst and O. Maimon}, title = {A Distributed Genetic Algorithm for Neural Network Design and Training}, journal = {Complex Systems}, year = {1992}, month = {Oct}, volume = {6}, number = {5}, pages = {459-477}, abstract = {ABSTRACT A new approach for designing and training neural networks is developed using a distributed genetic algorithm. A search for the optimal architecture and weights of a neural network comprising binary, linear threshold units is performed. For each individual unit, the Authors look "o" the optimal set of connections and associated weights under the restriction of a feedforward network structure. This is accomplished with the modified genetic algorithm, using an objective function-fitness-that considers, primarily, the overall network error; and, secondarily, using the unit's possible connections and weights that are preferable for continuity of the convergence process. Examples are given showing the potential of the proposed approach.}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = { } } @inproceedings{Oosthuizen89, author = {Oosthuizen, G.D.}, title = {Machine Learning: A Mathematical Framework for Neural Network, Symbolic and Genetic-Based Learning}, booktitle = {Proceedings of the Third International Conference on Genetic Algorithms}, year = {1989}, pages = {385-390}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Opitz94, author = {D. W. Opitz and J. W. Shavlik}, title = {Genetically Refining Topologies of Knowledge-Based Neural Networks}, booktitle = {International Symposium on Integrating Knowledge and Neural Heuristics}, year = {1994}, month = {May}, address = {Pensacola, FL}, pages = {57-66}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Opitz94a, author = {D. W. Opitz and J. W. Shavlik}, title = {Using Genetic Search to Refine Knowledge-Based Neural Networks}, booktitle = {Machine Learning: Proceedings of the Eleventh International Conference}, year = {1994}, month = {July}, publisher = {Morgan Kaufmann}, address = {New Brunswick, NJ}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @incollection{Paredis90, key = {genetic algorithms, connectionism cogann}, author = {Jan Paredis}, title = {The Evolution of Behavior: Some Experiments}, booktitle = {Simulation of Adaptive Behavior}, year = {1990} } @article{Parisi90, author = {Parisi, D. and Cecconi, F. and Nolfi, S.}, title = {Econets: Neural Networks That Learn in an Environment}, journal = {Network}, year = {1990}, volume = {1}, pages = {149-168}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Parisi91, author = {Parisi, D. and Nolfi, S. and Cecconi, F.}, title = {Learning, Behaviour and Evolution}, institution = {C.N.R. Rome}, year = {1991}, number = {PCIA-91-14}, note = {Also in Proceedings of the first european conference on artificial life ECAL 91, pp.207-216, Varela,F.J. and Bourgine,P. (Eds)}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Penfold93a, author = {Penfold, H.B. and Kohlmorgen, U. and Schmeck, H.}, title = {Deriving Application-Specific Neural Nets Using a Massively Parallel Genetic Algorithm}, institution = {The University of Newcastle, Australia}, year = {1993}, number = {Parallel GA 004}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Petridis93, author = {V. Petridis and S. Kazarlis and A. Papaikonomou}, title = {A Genetic Algorithm for Training Recurrent Neural Networks}, journal = {Proceedings of IJCNN '93, Nagoya Japan}, year = {1993}, pages = {2706-2709}, topology = {recurrent}, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @article{Petridis92, author = {V. Petridis and S. Kazarlis and A. Papaikonomou and A. Filelis}, title = {A Hybrid Genetic Algorithm for Training Neural Networks}, journal = {proceedings of ICANN '92, Brighton England}, year = {1992}, pages = {953-956}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @inproceedings{Philipsen93, author = {Philipsen, W.J.M. and Cluitmans, L.J.M.}, title = {Using a genetic algorithm to tune Potts neural networks}, booktitle = {Artificial Neural Nets and Genetic Algorithms}, year = {1993}, editor = {Albrecht, R.F. and Reeves, C.R. and Steele, N.C.}, publisher = {Springer-Verlag}, pages = {650-657}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Porto90, key = {genetic algorithms, connectionism, cogann ref}, author = {Vincent W. Porto and David B. Fogel}, title = {Neural Network Techniques for Navigation of AUVs}, booktitle = {Proceedings of the IEEE Symposium on Autonomous Underwater Vehicle Technology}, year = {1990}, month = {5-6 Jun}, address = {Washington, DC}, pages = {137-141} } @inproceedings{Potter92, author = {Potter, M.A.}, title = {A Genetic Cascade-Correlation Learning Algorithm}, booktitle = {Proceedings of COGANN-92 International Workshop on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, editor = {Whitley, L.D. and Schaffer, J.D.}, publisher = {IEEE Computer Society Press}, pages = {123-133}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Prados92, key = {genetic algorithms connectionism}, author = {Donald L. Prados}, title = {New Learning Algorithm for Training Multilayered Neural Networks that Uses Genetic-Algorithm Techniques}, journal = {Electronics Letters}, year = {1992}, month = {July}, volume = {28}, number = {16}, pages = {1560-1561}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @article{Prados92a, key = {genetic algorithms connectionism neural networks cogann}, author = {Donald L. Prados}, title = {Training Multilayered Neural Networks by Replacing the Least Fit Hidden Neurons}, journal = {Proceedings of IEEE SOUTHEASTCON}, year = {1992}, volume = {2}, pages = {634-637}, publisher = {IEEE}, address = {Piscataway, NJ}, abstract = {ABSTRACT The Author discusses " supervised-learning algorithm, called GenLearn, for training multilayered neural networks. GenLearn uses techniques from the field of genetic algorithms to perform a global search of weight space and, thereby, to avoid the common problem of getting stuck in local minima. GenLearn is based on survival of the fittest hidden neuron. In searching for the most fit hidden neurons, GenLearn searches for a globally optimal internal representation of the input data. A big advantage of the GenLearn procedure over the generalized delta rule (GDR) in training three-layered neural nets is that, during each iteration of GenLearn, each weight in the first matrix is modified only once, whereas, in the GDR procedure, each weight in the first matrix is modified once for each output-layer neuron. What makes this such a big advantage is that, although GenLearn often reaches the desired mean square error in about the same number of iterations as the GDR, each iteration takes considerably less time.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @techreport{Radcliffe90, key = {genetic algorithms, connectionism, cogann ref}, author = {Nicholas J. Radcliffe}, title = {Genetic Neural Networks on MIMD Computers}, institution = {Dept. of Theoretical Physics University of Edinburgh}, year = {1990}, address = {Edinburgh, Scotland}, type = {Ph.D. DISS}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Reeves92, key = {genetic algorithms connectionism neural networks cogann}, author = {C. Reeves and N. Steele}, title = {Problem-Solving by Simulated Genetic Processes: A Review and Application to Neural Networks}, booktitle = {Proceedings of the Tenth IASTED International Conference. Applied Informatics}, year = {1992}, editor = {M.H. Hamza}, publisher = {Acta Press}, address = {Zurich, Switzerland}, pages = {269-272}, abstract = {ABSTRACT In the past decade, researchers have become aware of the value of simulating natural processes in order to solve large and difficult problems. One example which is attracting increasing attention is the idea of a genetic algorithm (GA). The first part of this paper provides a review of the basic concepts underlying genetic algorithms. The methodology is illustrated by a simple example, and some of the issues involved in more advanced GAs are discussed. Finally, it describes some of their applications. The second part describes in some detail research carried out in applying genetic algorithms to the field of neural networks, in particular to the multi-layer perceptron (MLP). This work falls into two main areas. The first of these deals with the question of the design of a neural network architecture, and the choice of a training regime for a particular problem. The second area of application is to the basic learning process itself. Traditionally, the MLP has been trained by a process called back-propagation. This paper reports on an alternative method based on a GA, and it is argued that such an approach has many advantages over back-propagation.}, topology = {multi-layered}, network = {multi-layer perceptron}, encoding = { }, evolves = {connectivity, parameters}, applications = { } } @article{Rehm92, key = {genetic algorithms connectionism neural networks cogann}, author = {W. Rehm and V. Sterzing}, title = {An Optimization Method for Multilayer Perceptron Based on Evolution-Theoretic Principles}, journal = {Informationstechnik - IT}, year = {1992}, month = {Oct}, volume = {34}, number = {5}, pages = {307-312}, abstract = {ABSTRACT Pattern classification tasks can be addressed successfully using multilayer perceptrons, i.e. a simple form of feed-forward neural nets. The training of a neural net can be regarded to be a parametric optimization problem, for which several possible algorithms are known. These differ in efficiency, given a fixed complexity of structure and classification task, and in the implementation constrains on parallel hardware. The Authors introduce " new method based on evolution-theoretic principles using operators from genetic algorithms, that is well suited for a parallel implementation on MIMD architectures. Finally they provide some results on parity problems.}, topology = {multi-layered}, network = {multi-layer perceptron}, encoding = { }, evolves = {parameters}, applications = { } } @inproceedings{Robbins93, key = {genetic algorithms connectionism neural networks cogann}, author = {P. Robbins and A. Soper and K. Rennolls}, title = {Use of Genetic Algorithms for Optimal Topology Determination in Back Propagation Neural Networks}, booktitle = {Proceedings of the International Conference on Artificial Neural Networks and Genetic Algorithms}, year = {1993}, pages = {726-730}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Romaniuk93, key = {genetic algorithms connectionism neural networks cogann}, author = {Steve G. Romaniuk}, title = {Evolutionary Growth Perceptrons}, booktitle = {Proceedings of the Fifth International Conference on Genetic Algorithms}, year = {1993}, network = {perceptron}, encoding = { }, evolves = { }, applications = { } } @techreport{Rudnick90, key = {connectionism, genetic algorithms, cogann ref}, author = {Michael Rudnick}, title = {A Bibliography: The Intersection of Genetic Search and Artificial Neural Networks}, institution = {Department of Computer Science and Engineering, Oregon Graduate Institute}, year = {1990}, type = {CS/E 90-001} } @techreport{Rudnick92, key = {connectionism, contiguity problem}, author = {Michael Rudnick}, title = {Genetic Algorithms and Fitness Variance with an Application to the Automatic Design of Artificial Neural Networks}, institution = {Oregon Graduate Institute of Science and Technology}, year = {1992}, address = {Portland, OR}, type = {Unpublished Ph.D. DISS}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @conference{Rudnick93, key = {connectionism, genetic algorithms, cogann}, author = {Michael Rudnick}, title = {Evolutionary Network Design and the Contiguity Problem}, booktitle = {Proceedings of the World Congress on Neural Networks 93}, organization = {WCNN93}, year = {1993}, pages = {IV135 - IV138}, abstract = {ABSTRACT Given a particular problem to solve using an artificial neural network, we wish to find a superior network architecture; this is called the network design problem. One approach is to use evolutionary methods, or evolutionary network design (END). The contiguity problem consists of counting the number of clumps of 1's in a binary input field. It is a good test problem for END because, for back-propagation networks, the space of network architectures has been characterized with respect to network generalization ability. We present experience gained using END to find superior network architectures for the contiguity problem.}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = {contiguity} } @inproceedings{Schaffer92, key = {genetic algorithms, connectionism, neural networks}, author = {J. David Schaffer and Darrell Whitley and Larry J. Eshelman}, title = {Combinations of Genetic Algorithms and Neural Networks: A Survey of the State of the Art}, booktitle = {Proceedings of the Conference on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, pages = {1-37}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Scherf92, key = {genetic algorithms connectionism neural networks cogann}, author = {A.V. Scherf and L.D. Voelz}, title = {Training Neural Networks with Genetic Algorithms for Target Detection}, journal = {Proceedings of the SPIE - The International Society for Optical Engineering}, year = {1992}, volume = {1710, pt.1}, pages = {II-734--II-41}, abstract = {ABSTRACT Algorithms for training artificial neural networks, such as backpropagation, often employ some form of gradient descent in their search for an optimal weight set. The problem with such algorithms is their tendency to converge to local minima, or not to converge at all. Genetic algorithms simulate evolutionary operators in their search for optimality. The techniques of genetic search are applied to training a neural network for target detection in infrared imagery. The algorithm design, parameters, and experimental results are detailed. Testing verifies that genetic algorithms are a useful and effective approach for neural network training.}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @inproceedings{Schiffmann93, key = {genetic algorithms connectionism neural networks cogann application classification of thyroid tests}, author = {W. Schiffmann and M. Joost and R. Werner}, title = {Application of Genetic Algorithms to the Construction of Topologies for Multilayer Perceptrons}, booktitle = {Proceedings of the International Conference on Artificial Neural Networks and Genetic Algorithms}, year = {1993}, pages = {675-682}, topology = {multi-layered}, network = {multi-layer perceptron}, encoding = { }, evolves = {connectivity}, applications = { } } @incollection{Schiffmann90, key = {connectionism}, author = {W. Schiffmann and K. Mecklenburg}, title = {Genetic Generation of Backpropagation Trained Neural Networks}, booktitle = {Parallel Processing in Neural Systems and Computers}, year = {1990}, editor = {R. Eckmiller and G. Hartmann and G. Hauske}, publisher = {Elsevier Science Publishers}, pages = {205-208}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Schizas92, key = {genetic algorithms, connectionism, neural networks}, author = {C.N. Schizas and C.S. Pattichis and L.T. Middleton}, title = {Neural Networks, Genetic Algorithms and K-Means Algorithm: In Search of Data Classification}, booktitle = {Proceedings of the Conference on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, pages = {201-222}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Sebald90, author = {Sebald, A.V. and Fogel, D.B.}, title = {Design of SLAYR Neural Networks Using Evolutionary Programming}, booktitle = {Proceedings of the Twenty-Fourth Asilomar Conference on Signals, Systems and Computers}, year = {1990}, editor = {Chen, R. R.}, publisher = {The Computer Society of IEEE/Maple Press}, pages = {1020-1024}, volume = {2}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Sebald91, key = {genetic algorithms, connectionism, cogann ref}, author = {A.V. Sebald and D.B. Fogel}, title = {Using Evolutionary Neural Networks for Arterial Waveform Discriminiation}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1991}, pages = {A-955}, volume = {II}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Sebald92, key = {genetic algorithms connectionism neural networks cogann}, author = {A.V. Sebald and J. Schlenzig and D.B. Fogel}, title = {Minimax Design of CMAC Encoded Neural Network Controllers Using Evolutionary Programming}, journal = {Asilomar Conference on Circuits, Systems \& Computers}, year = {1992}, volume = {1}, pages = {551-555}, publisher = {Maple Press, Inc}, address = {San Jose, CA, USA}, abstract = {ABSTRACT The Authors describe "h" use of evolutionary programming for computer-aided design and testing of cerebellar model arithmetic computer (CMAC) encoded neural network regulators. The design and testing problem is viewed as a game in that the controller parameters are to be chosen with a minimax criterion, i.e. to minimize the loss associated with their use on the worst possible plant parameters. The technique permits analysis of neural strategies against a set of plants. This gives both the best choice of control parameters and identification of the plant configuration which is most difficult for the best controller to handle.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Sharp91, key = {connectionism, cogann ref}, author = {David H. Sharp and John Reinitz and Eric Mjolsness}, title = {Genetic Algorithms for Genetic Neural Nets}, institution = {Department of Computer Science, Yale University}, year = {1991}, month = {Jan}, type = {Research Report YALEU/DCS/TR-845}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Shibata93, author = {Shibata, T. and Fukada, T. and Tanie, K.}, title = {Nonlinear Backlash Compensation Using Recurrent Neural Network - Unsupervised Learning by Genetic Algorithm}, booktitle = {IJCNN'93-NAGOYA Proceedings of the 1993 International Joint Conference on Neural Networks, Nagoya (Japan)}, year = {1993}, publisher = {IEEE}, pages = {742-745}, topology = {recurrent}, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @inproceedings{Shibata93a, author = {Shibata, T. and Fukada, T. and Tanie, K.}, title = {Fuzzy Critic for Robotic Motion Planning by Genetic Algorithm in Hierarchical Intelligent Control}, booktitle = {IJCNN'93-NAGOYA Proceedings of the 1993 International Joint Conference on Neural Networks, Nagoya (Japan)}, year = {1993}, publisher = {IEEE}, pages = {770-773}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {robot controller} } @inproceedings{Shibata93b, author = {Shibata, T. and Fukada, T. and Tanie, K.}, title = {Synthesis of Fuzzy Artificial Intelligence, Neural Networks, and Genetic Algorithms for Heirarchical Intelligent Control}, booktitle = {IJCNN'93-NAGOYA Proceedings of the 1993 International Joint Conference on Neural Networks, Nagoya (Japan)}, year = {1993}, publisher = {IEEE}, pages = {2869-2872}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Shonkwiler92, key = {genetic algorithms, connectionism, neural networks}, author = {R. Shonkwiler and Kenyon R. Miller}, title = {Genetic Algorithm/Neural Network Synergy For Nonlinear Constrained Optimization Problems}, booktitle = {Proceedings of the Conference on Combinations of Genetic Algorithms and Neural Networks}, year = {1992}, pages = {248-257}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {optimization} } @techreport{Smieja92, key = {genetic algorithms connectionism neural networks cogann}, author = {F.J. Smieja}, title = {Evolution of Intelligent Systems in a Changing Environment}, institution = {Gesellschaft fuer Mathematik und Datenverarbeitung m.b.H.}, year = {1992}, address = {Bonn, Germany}, type = {GMD-623}, pages = {24}, abstract = {ABSTRACT In the report a first version of a simulated robot is described, which will embody both neural network and genetic algorithm optimization procedures. The system is modularly structured, with neural networks at the lower (recognition) level of the simple brain of the robot, and at the higher level prescribed decision behaviors are followed. It is the higher level parameters determining the nature of the decisions made that are to be optimized via genetic algorithms. Having sketched the structure and method of operation of the prototype robot, a community of robots situation is introduced as the next stage, for optimization within a robot-inhabited world.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {simulated world} } @article{Smith94, author = {Smith, R. E. and Cribbs, H. B.}, title = {Is a Learning Classifier System a Type of Neural Network?}, journal = {Evolutionary Computation}, year = {1994}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Smith93, key = {cogann genetic algorithms, connectionism, neural networks, classifier systems}, author = {Robert E. Smith}, title = {Genetic Learning in Rule-Based and Neural Systems}, journal = {Proceedings of the Third International Workshop on Neural Networks and Fuzzy Logic}, year = {1993}, month = {Jan}, volume = {1}, pages = {183}, publisher = {NASA. Johnson Space Center}, abstract = {ABSTRACT The design of neural networks and fuzzy systems can involve complex, nonlinear, and ill-conditioned optimization problems. Often, traditional optimization schemes are inadequate or inapplicable for such tasks. Genetic Algorithms (GA's) are a class of optimization procedures whose mechanics are based on those of natural genetics. Mathematical arguments show how GAs bring substantial computational leverage to search problems, without requiring the mathematical characteristics often necessary for traditional optimization schemes (e.g., modality, continuity, availability of derivative information, etc.). GA's have proven effective in a variety of search tasks that arise in neural networks and fuzzy systems. This presentation begins by introducing the mechanism and theoretical underpinnings of GA's. GA's are then related to a class of rule-based machine learning systems called learning classifier systems (LCS's). An LCS implements a low-level production-system that uses a GA as its primary rule discovery mechanism. This presentation illustrates how, despite its rule-based framework, an LCS can be thought of as a competitive neural network. Neural network simulator code for an LCS is presented. In this context, the GA is doing more than optimizing and objective function. It is searching for an ecology of hidden nodes with limited connectivity. The GA attempts to evolve this ecology such that effective neural network performance results. The GA is particularly well adapted to this task, given its naturally-inspired basis. The LCS/neural network analogy extends itself to other, more traditional neural networks. Conclusions to the presentation discuss the implications of using GA's in ecological search problems that arise in neural and fuzzy systems.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Spears90, key = {COGANN connectionism}, author = {W.M. Spears and K.A. De Jong}, title = {Using Neural Networks and Genetic Algorithms as Heuristics for NP-Complete Problems}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1990}, pages = {118 - 121}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Spiessens92, key = {genetic algorithms connectionism neural networks cogann}, author = {P. Spiessens and J. Torreele}, title = {Massively Parallel Evolution of Recurrent Networks: An Approach to Temporal Processing}, booktitle = {Toward a Practice of Autonomous Systems. Proceedings of the First European Conference on Artificial Life}, year = {1992}, editor = {F.J. Varela and P. Bourgine}, publisher = {MIT Press}, address = {Cambridge, MA, USA}, pages = {70-77}, abstract = {ABSTRACT The Authors investigate "" evolutionary approach to the problem of time-dependent processing with recurrent networks. Both structure and weights of these networks are evolved by a fine-grained parallel genetic algorithm. The parallel nature of this algorithm, which enables the co-evolution of clusters of networks, made it possible to successfully solve three non-trivial temporal processing problems. One of these problems consists of evolving a trail-following behaviour for an artificial ant.}, topology = {recurrent}, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = {simulated world} } @inproceedings{Stacey91, key = {genetic algorithms, connectionism, cogann ref}, author = {Deborah A. Stacey and Stefan Kremer}, title = {The Guelph Darwin Project: The Evolution of Neural Networks by Genetic Algorithms}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1991}, pages = {A-957}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { }, journal = {IJCNN-91}, volume = {II} } @inproceedings{Stork90, key = {genetic algorithms, connectionism}, author = {D.G. Stork and S. Walker and M. Burns and B. Jackson}, title = {Preadaptation in Neural Circuits}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1990}, pages = {I-202 - I-205}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Suzuki91, key = {genetic algorithms applications pattern classification categorization; relation AI machine learning connectionist networks basin; associative memory model; analysis, neural, cogann ref}, author = {Keiji Suzuki and Yukinori Kakazu}, title = {An Approach to the Analysis of the Basins of the Associative Memory Model Using Genetic Algorithms}, booktitle = {Proceedings of the Fourth International Conference on Genetic Algorithms}, year = {1991}, pages = {539-546}, abstract = {Abstract: In this paper, an approach to the analysis of the brain of a correlational associative memory model using the Genetic Algorithms and a new training algorithm for this model is described. The recalling process of a model described by direction cosine is insufficient for the better understanding of the dynamical behavior of the model. In order to know the characteristics of memorized states, the methodology of the Genetic Algorithms applied to analyze the recalling process concerned with each memorized state is proposed. Furthermore, before the analyzing, the LU-algorithm is proposed to give the model the ability of keeping a wide basin in both highly memorized rates and mutually non-orthogonal states. Finally, results of experiments related to the basin analysis are shown.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Tackett91, key = {connectionism, genetic algorithms, cogann}, author = {Walter Alden Tackett}, title = {Genetic Generation of Dendritic Trees for Image Classification}, booktitle = {Proceedings of the World Congress on Neural Networks}, year = {1993}, pages = {IV646 - IV649}, abstract = {ABSTRACT Genetic Programming (GP) is an adaptive method for generating executable programs from labeled training data. It differs from the conventional methods of Genetic Algorithms because it manipulates tree structures of arbitrary size and shape rather than fixed length binary strings. We apply GP to the development of a processing tree with a dendritic, or neuron-like structure: measurements from a set of input nodes are weighted and combined through linear and nonlinear operations to form an output response. Unlike conventional neural methods, no constraints are placed upon size, shape, or order of processing withing the network. This network is used to classify feature vectors extracted from IR imagery into target/nontarget catagories using a database of 2000 training samples. Performance is tested against a separate database of 7000 samples. For purposes of comparison, the same training and test sets are used to train two other adaptive classifier systems, the binary tree classifier and the Backpropagation neural network. The GP network acheives higher performance with reduced computational requirements.}, topology = { }, network = { }, encoding = {indirect, LISP program}, evolves = { }, applications = {image classification} } @inproceedings{Takagi93, key = {connectionism, genetic algorithms, cogann}, author = {Hideyuki Takagi}, title = {Neural Network and Genetic Algorithm Techniques for Fuzzy Systems}, booktitle = {Proceedings of the World Congress on Neural Networks}, year = {1993}, pages = {II631 - II634}, abstract = {ABSTRACT This paper introduces (1) how neural networks and genetic algorithms have been used for auto-designing fuzzy systems, (2) how neural networks are combined with fuzzy systems in commecial applications, and (3) how fuzzy systems are used to improve the performance of neural networks and genetic algorithms.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Tamburino92, key = {genetic algorithms connectionism neural networks cogann cooperative application}, author = {Louis A. Tamburino and Mateen M. Rizki}, title = {Performance-Driven Autonomous Design of Pattern-Recognition Systems}, journal = {Applied Artificial Intelligence}, year = {1992}, volume = {6}, number = {1}, pages = {59-77}, abstract = {ABSTRACT The closed-loop design experiment described in this paper demonstrates a three-phase automated design approach to pattern recognition. The experiment generates morphological feature detectors and then uses a novel application of genetic algorithms to select cooperative sets of features to pass to a neural net classifier. The self-organizing hybrid learning approach embodied in this closed-loop design methodology is complementary to conventional artificial intelligence (AI) expert systems that utilize rule-based approaches and a specific set of design elements. This experiment is part of a study directed to emulating the nondirected processes of biological evolution. The approach we discuss is semiautomatic in that initialization of computer programs requires human experience and expertise to select representations, develop search strategies, choose performance measures, and devise resource-allocation strategies. The hope is that these tasks will become easier with experience and will provide the means to exploit parallel processing without the need to analyze or program an entire design solution.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Thierens93, key = {genetic algorithms connectionism neural networks cogann pole inversion}, author = {D. Thierens and J. Suykens and J. Vandewalle and B. De Noor}, title = {Genetic Weight Optimization of a Feedforward Neural Network Controller}, booktitle = {Artificial Neural Nets and Genetic Algorithms Proceedings of the International Conference at Innsbruck, Austria}, year = {1993}, editor = {R.F.~Albrecht and C.R.~Reeves and N.C.~Steele}, publisher = {Springer}, address = {Wien and New York}, pages = {658-663}, abstract = {ABSTRACT The optimization of the weights of a feedforward neural network with a genetic algorithm is discussed. The search by the recombination operator is hampered by the existence of two functionally equivalent symmetries in feedforward neural networks. To sidestep these representation redundancies we reorder the hidden neurons on the genotype before recombination according to a weight sign matching criterion, and flip the weight signs of a hidden neuron's connections whenever there are more inhibitory than excitatory incoming and outgoing links. As an example we optimize a feedforward neural network that implements a nonlinear optimal control law. The neural controller has to swing up the inverted pendulum from its lower equilibrium point to its upper equilibrium point and stabilize it there. Finding weights of the network represents a nonlinear optimization problem which is solved by the genetic algorithm.}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {parameters}, applications = {controller} } @unpublished{Todd88, author = {Todd, P.}, title = {Evolutionary Methods for Connectionist Architectures}, year = {1988}, note = {unpublished internal report, Stanford}, institution = {Stanford}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Todd??, key = {genetic algorithm, connectionism, Hebbian learning}, author = {Peter M. Todd and Geoffery F. Miller}, title = {Exploring Adaptive Agency II: Simulating the Evolution of Associative Learning}, booktitle = {Proceedings of the International Conference on Simulation of Adaptive Behavior: From Animals to Animats}, editor = {S. Wilson and J.-A. Meyer}, publisher = {MIT Press}, pages = {306-315}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Torreele91, author = {Torreele, J.}, title = {Temporal Processing With Recurrent Networks: An Evolutionary Approach}, booktitle = {Proceedings of the Fourth International Conference on Genetic Algorithms}, year = {1991}, editor = {Belew, R.K. and Booker, L.B.}, publisher = {Morgan Kaufmann}, pages = {555-561}, abstract = {Abstract: In this paper we present an evolutionary approach to the problem of temporal processing with recurrent networks. A genetic algorithm is used to evolve both structure and weights, so as to alleviate the design and learning problem recurrent networks suffer from. The viability of this approach is demonstrated by successfully solving two nontrivial temporal processing problems. The important technique of teacher forcing is identified and its influence on the performance of the algorithm is empirically demonstrated.}, topology = {recurrent}, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = {temporal pattern recognition} } @conference{Toth93, key = {connectionism, genetic algorithms, cogann}, author = {Gabor J. Toth and Andras Lorincz}, title = {Genetic Algorithm With Migration on Topology Conserving Maps}, booktitle = {Proceedings of the World Congress on Neural Networks}, organization = {WCNN93}, year = {1993}, pages = {III168 - III171}, abstract = {ABSTRACT Optimization problems depending on external variables (parameters) are treated with the help of a Kohonen network extended by a genetic algorithm (GA). The optimal solution is assumed to have continuous dependence on the external variables. The GA was generalized to organize individuals into subpopulations, which were allocated in the space of the external variables in an optimal fashion by Kohnonen digitization. Individuals were allowed to breed within their own subpopulations and in neighboring ones (migration). To illustrate the strength of the modified GA the optimal control of a simulated robot-arm is treated: a falling ping-pong ball has to be caught by a bat without bouncing. It is shown that the simultaneous optimization problem (for different values of the external parameter) can be solved successfully, and the migration can considerably reduce computation time.}, network = {kohonen}, encoding = { }, evolves = { }, applications = { } } @incollection{Uhr63, author = {Uhr, L. and Vossler, C.}, title = {A Pattern Recognition Program that Generates, Evaluates and Adjusts its Own Operators}, booktitle = {Computers and Thought}, year = {1963}, editor = {Feigenbaum, E. and Feldman, J.}, publisher = {McGraw-Hill, New York}, topology = {feed-forward}, network = { }, encoding = { }, evolves = {feature detectors}, applications = {pattern classification} } @inproceedings{Uthmann93, key = {connectionism cogann}, author = {T. Uthmann and D. Polani}, title = {Training Kohonen Feature Maps in Different Topologies: An analysis using Genetic Algorithms}, booktitle = {Proceedings of the International Conference on Genetic Algorithms}, year = {1993}, network = {kohonen}, encoding = { }, evolves = { }, applications = { } } @phdthesis{Vaario93, author = {Jari Vaario}, title = {An Emergent Modeling Method for Artificial Neural Networks}, year = {1993}, school = {The University of Tokyo}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Vaario94, author = {Jari Vaario}, title = {Artificial Life as Constructivist {AI}}, journal = {Journal of {SICE} ({J}apanese {S}ociety of {I}nstrument and {C}ontrol {E}ngineers)}, year = {1994}, volume = {33}, number = {1}, pages = {65-71}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Vaario94a, author = {Jari Vaario}, title = {From Evolutionary Computation to Computational Evolution}, journal = {Informatica}, year = {1994}, note = {(to appear)}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Vaario94b, author = {Jari Vaario}, title = {Modeling Adaptative Self-Organization}, booktitle = {Proceedings of Artificial {L}ife {IV}}, year = {1994}, month = {July 6-8}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Vaario93a, author = {Jari Vaario and Koichi Hori and Setsuo Ohsuga}, title = {Toward Evolutionary Design of Autonomous Systems}, journal = {The International Journal in Computer Simulation}, year = {1994}, note = {(to appear)}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @incollection{Vaario91, author = {Jari Vaario and Setsuo Ohsuga}, title = {Adaptive Neural Architectures through Growth Control}, booktitle = {Intelligent Engineering Systems through Artificial Neural Networks}, year = {1991}, pages = {11--16}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Vaario92, author = {Jari Vaario and Setsuo Ohsuga}, title = {An Emergent Construction of Adaptive Neural Architectures}, journal = {Heuristics - The Journal of Knowledge Engineering}, year = {1992}, volume = {5}, number = {2}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @incollection{Vaario93b, author = {Jari Vaario and Setsuo Ohsuga}, title = {On Growing Intelligence}, booktitle = {Neural Networks and a New {AI}}, year = {1994}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @incollection{Vaario91a, author = {Jari Vaario and Setsuo Ohsuga and Koichi Hori}, title = {Connectionist Modeling Using {Lindenmayer} Systems}, booktitle = {Information Modeling and Knowledge Bases: Foundations, Theory, and Applications}, year = {1991}, pages = {496--510}, topology = { }, network = { }, encoding = {indirect, L-systems}, evolves = { }, applications = { } } @article{Vemuri92, author = {V. Vemuri}, title = {Neural Networks Can be Used for Open-Loop, Dynamic Control}, journal = {International Journal of Neural Networks}, year = {1992}, volume = {2}, pages = {71-84}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {control systems} } @inproceedings{Vico91, author = {Vico, F.J. and Sandoval, F.}, title = {Use of Genetic Algorithms in Neural Networks Definition}, booktitle = {Artificial Neural Networks, IWANN91, Granada}, year = {1991}, editor = {Prieto, A.}, publisher = {Lecture notes in Computer Science 540, Springer Verlag}, pages = {196-203}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Vico92, key = {genetic algorithms connectionism neural networks cogann}, author = {F.J. Vico and F. Sandoval}, title = {Neural Networks Definition Algorithm}, journal = {Microprocessing and Microprogramming}, year = {1992}, volume = {34}, number = {1-5}, pages = {251-254}, abstract = {ABSTRACT There is not a general methodology for neural network definition. The Authors propose "" algorithm highly inspired on biological concepts for generating neural networks oriented to solve particular problems given on terms of input and output. With this algorithm they intend to specify formal tools of general use for network definition, and to disclose underlying processing structures of the living organisms. The concepts of genetic code, embryogenesis and evolution are the main keys in the development of the algorithm they propose.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Voigt93, author = {Voigt, H-M. and Born, J. and Santibanez-Koref, I.}, title = {Evolutionary Structuring of Artificial Neural Networks}, institution = {Bionics and Evolution Techniques Laboratory, Technical University Berlin}, year = {1993}, number = {TR-02-93}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Weiss90, author = {Weiss, G.}, title = {Combining Neural and Evolutionary Learning: Aspects and Approaches}, institution = {Technical University of Munich}, year = {1990}, month = {May}, type = {TUM FKI-132-90}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Werner90, author = {Werner, G.M. and Dyer, M.G.}, title = {Evolution of Communication in Artificial Organisms}, institution = {AI Lab, UCSD}, year = {1990}, number = {UCLA-AI-90-06}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {simulated world} } @inproceedings{Werner91, key = {genetic algorithm connectionism neural networks}, author = {G.M. Werner and M.G. Dyer}, title = {Evolution of Communication in Artificial Organisms}, booktitle = {Artificial Life II: Proceedings of the Workshop on Artificial Life Held in 1990}, year = {1991}, pages = {659-687}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {simulated world} } @article{Whitaker93, key = {genetic algorithms connectionism neural networks cogann}, author = {Kevin W. Whitaker and Ravi K. Prasanth and Robert E. Markin}, title = {Specifying Exhaust Nozzle Contours With a Neural Network}, journal = {AIAA Journal}, year = {1993}, month = {Feb}, volume = {31}, number = {2}, pages = {273-277}, abstract = {ABSTRACT Thrust vectoring is continuing to become an important issue in future military aircraft system designs. A recently developed concept of vectoring aircraft thrust makes use of flexible exhaust nozzles. Subtle modifications in the nozzle wall contours produce a nonuniform flowfield containing a complex pattern of shock and expansion waves. The end result, due to the asymmetric velocity and pressure distributions, is vectored thrust. Specification of the nozzle contours required for a desired thrust vector angle (an inverse design problem) has been achieved with genetic algorithms. However, this approach is computationally intensive, preventing nozzles from being designed on demand, which is necessary for an operational aircraft system. An investigation was conducted into using genetic algorithms to train a neural network in an attempt to obtain, in real time, two-dimensional nozzle contours. Results show that genetic-algorithm trained neural networks provide a viable, time-efficient alternative for designing thrust vectoring nozzle contours. Thrust vector angles up to 20 deg were obtained within an average error of 0.0914 deg. The error surfaces encountered were highly degenerate and thus the robustness of genetic algorithms was well suited for minimizing global errors.}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @article{Whitehead94, author = {Whitehead, B.A. and Choate, T.D.}, title = {Evolving Space-Filling Curves to Distribute Radial Basis Functions Over an Input Space}, journal = {IEEE Transactions on Neural Networks}, year = {1994}, volume = {5}, pages = {15-23}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Whitley90a, author = {Whitley, D. and Bogart, C.}, title = {The Evolution of Connectivity: Pruning Neural Networks Using Genetic Algorithms}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1990}, publisher = {IEEE Press}, pages = {134-137}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Whitley91, author = {Whitley, D. and Dominic, S. and Das, R.}, title = {Genetic Reinforcement Learning with Multilayer Neural Networks}, booktitle = {Proceedings of the Fourth International Conference on Genetic Algorithms}, year = {1991}, editor = {Belew, R. K. and Booker, L. B.}, publisher = {Morgan Kaufmann, San Mateo, CA}, pages = {562-569}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Whitley89, author = {Whitley, D. and Hanson, T.}, title = {Optimizing Neural Networks Using Faster, More Accurate Genetic Search}, booktitle = {Proceedings of the Third International Conference on Genetic Algorithms}, year = {1989}, editor = {Schaffer, J.D.}, publisher = {Morgan Kaufmann}, pages = {391-396}, institution = {Computer Science Dept. Colorado Univ}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Whitley90, author = {Whitley, D. and Starkweather, T. and Bogart, C.}, title = {Genetic Algorithms and Neural Networks: Optimizing Connections and Connectivity}, journal = {Parallel Computing}, year = {1990}, volume = {14-3}, pages = {347-361}, institution = {Colorado State University}, topology = { }, network = { }, encoding = { }, evolves = {connectivity, parameters}, applications = { } } @inproceedings{Whitley88, key = {connectionism, cogann ref}, author = {Darrell Whitley}, title = {Applying Genetic Algorithms to Neural Network Problems: A Preliminary Report}, booktitle = {Proceedings of the International Neural Network Society Conference}, organization = {PROC INNS88}, year = {1988}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Whitley88a, key = {connectionism, cogann ref}, author = {Darrell Whitley}, title = {Applying Genetic Algorithms to Neural Network Learning}, institution = {Department of Computer Science, Colorado State University}, year = {1988}, number = {CS-88-128}, note = {Also appeared in: Proceedings of the 7th Conference for the Study of Artificial Intelligence and Simulated Behavior, Sussex, England 1989. Pitman Publishers.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @book{Whitley89a, key = {cogann ref, connectionism}, author = {Darrell Whitley}, title = {Optimizing Neural Networks Using Genetic Algorithms}, year = {1989}, publisher = {Markt and Technik}, address = {Munich, Germany}, journal = {Special Neurocomputing Issue of Design and Electronik}, topology = { }, network = { }, encoding = { }, evolves = {parameters}, applications = { } } @inproceedings{Whitley89b, key = {connectionism, cogann ref}, author = {Darrell Whitley}, title = {Genetic Algorithm Applications: Neural Nets, Traveling Salesmen and Schedules}, booktitle = {Proceedings of the 1989 Rocky Mountain Conference on Artificial Intelligence}, year = {1989}, address = {Denver, CO}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @conference{Whitley92, key = {genetic algorithms, connectionism, hill-climbing, mutation only, cogann ref}, author = {Darrell Whitley and S. Dominic and R. Das and C. Anderson}, title = {Genetic Reinforcement Learning for Neurocontrol Problems}, organization = {Machine Learning}, year = {1992}, abstract = {Abstract Empirical tests indicate that the class of genetic algorithms which have been shown to yield good performance for neural network weight optimization are really genetic hill-climbers, with a strong reliance on mutation rather than hyperplane sampling. These results are consistent with the theoretical results of Goldberg (1991) analyzing real-coded genetic algorithms. We argue that neural network learning applications such as neurocontrol problems are perhaps more appropriate for these genetic hill-climbers than supervised learning applications because in reinforcement learning applications gradient information is not directly available. On an inverted pendulum control problem reinforcement learning produces competitive results with AHC, another well-known reinforcement learning paradigm for neural networks that employs temporal difference methods. The genetic hill-climbing algorithm appears to be robust over a wide range of learning conditions. We also discuss several approaches for evaluating neural network performance.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Whitley, key = {genetic algorithms relation AI machine learning connectionist networks; reinforcement learning, connectionism, training, real coding, pole balancing, cogann ref}, author = {Darrell Whitley and Stephen Dominic and Rajarshi Das}, title = {Genetic Reinforcement Learning with Multilayer Neural Networks}, booktitle = {Proceedings of the International Conference on Genetic Algorithms}, year = {1991}, pages = {562-5769}, abstract = {Abstract: Empirical tests indicate that the genetic algorithms which have produced good performance for neural network weight optimization are really genetic hill-climbers, with a strong reliance on mutation rather than hyperplane sampling. Initial results are presented using genetic hill-climbers for reinforcement learning with multilayer neural networks for the control of a simulated cart-centering and pole-balancing dynamical system. "Genetic reinforcement learning" produces competitive results with AHC, a well-known reinforcement learning paradigm for neural networks that employs temporal difference methods.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Wieland90, author = {Alexis P. Wieland}, title = {Evolving Controls for Unstable Systems}, booktitle = {Proceedings of the 1990 Connectionist Models Summer School}, year = {1990}, editor = {Touretzky, D.S. and Elman, J.L. and Sejnowski, T.J. and Hinton, G.E.}, publisher = {Morgan Kaufman}, pages = {91-102}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Wieland91, key = {genetic algorithms, connectionism, pole balancing problems, cogann ref}, author = {Alexis P. Wieland}, title = {Evolving Neural Network Controllers for Unstable Systems}, journal = {IJCNN-91}, year = {1991}, volume = {II}, pages = {667-673}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Wilke93, key = {connectionism, genetic algorithms, cogann}, author = {Peter Wilke}, title = {Simulation of Neural Network and Genetic Algorithms in a Distributed Computing Environment Using NeuroGraph}, booktitle = {Proceedings of the World Congress on Neural Networks}, year = {1993}, pages = {I269 - I272}, abstract = {ABSTRACT NeuroGraph is a simulation environment for design, construction and execution of neural networks and genetic algorithms in a distributed computing environment. The simulator parts either run on single computers or as distributed applications on Unix/X-based networks, consisting of personal computers, workstations, or multi-processors. The parallelization component offers the possibility to divide computational tasks into concurrently executable modules, according to restrictions due to the neural net topology and computer net capabilities, ie. NeuroGraph tries to select the best configuration out of the available distributed hardware environment to fit performance requirements.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Williams94, author = {Williams, B. V. and Bostock, R. T. J. and Bounds, D. G. and Harget, A. J.}, title = {Improving Classification Performance in the Bumptree Network by Optimising Topology with a Genetic Algorithm}, booktitle = {IEEE Evolutionary Computation 1994}, year = {1994}, abstract = {ABSTRACT: The Bumptree is a binary tree of Gaussians which partitions a Euclidian space. The leaf layer consists of a set of local linear classifiers, and the whole system can be trained in a supervised manner to form a piecewise linear model. In this paper a Genetic Algorithm (GA) is used to optimise the topology of the tree. We discuss the properties of the genetic coding scheme, and argue that the GA/bumptree does not suffer from the same scaling problems as other GA/neural-net hybrids. Results on test problems, including a non-trivial classification task, are encouraging, with the GA able to discover topologies which give improved performance over those generated by a constructive algorithm.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Williams93, author = {Williams, B.V. and Bounds, D.G.}, title = {Learning and Evolution in Populations of Backprop Networks}, booktitle = {Proceedings of ECAL93 -- European Conference on Artificial Life}, year = {1993}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @misc{Wilson??, author = {C. L. Wilson and O.M. Omidvar}, title = {Optimization of Neural Network Topology and Information Content Using Boltzmann Methods}, year = {????}, howpublished = {NIST ir4766, at the NIST archive}, note = {available via ftp from (the NIST archive)}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @incollection{Wilson90, key = {genetic algorithm, connectionism, cogann ref}, author = {Stewart W. Wilson}, title = {Perceptron Redux: Emergence of Structure}, booktitle = {Emergent Computation}, year = {1990}, editor = {Stephanie Forrest}, publisher = {North Holland}, address = {Amsterdam}, pages = {249-256}, topology = { }, network = { }, encoding = { }, evolves = {connectivity}, applications = { } } @inproceedings{Yaeger94, author = {Larry Yaeger}, title = {Computational Genetics, Physiology, Metabolism, Neural Systems, Learning, Vision and Behavior or PolyWorld: Life in a New Context}, booktitle = {Artificial Life III, Proceedings Volume XVII}, organization = {Santa Fe Institute Studies in the Sciences of Complexity}, year = {1994}, editor = {C. G. Langton}, publisher = {Addison-Wesley}, pages = {263-298}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = {simuated world} } @inproceedings{Yamada93, author = {Yamada, T. and Yabuta, T.}, title = {Remarks on Neural Network Controller Which Uses Genetic Algorithm}, booktitle = {IJCNN'93-NAGOYA Proceedings of the 1993 International Joint Conference on Neural Networks, Nagoya (Japan)}, year = {1993}, publisher = {IEEE}, pages = {2783-2786}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @techreport{Yao92, author = {Xin Yao}, title = {A Review of Evolutionary Artificial Neural Networks}, institution = {Commonwealth Scientific and Industrial Research Organization, Division of Building, Construction and Engineering}, year = {1992}, address = {PO Box 56, Highett, Victoria 3190, Australia}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @article{Yao93, key = {genetic algorithms connectionism neural networks cogann}, author = {Xin Yao}, title = {A Review of Evolutionary Artificial Neural Networks}, journal = {International Journal of Intelligent Systems}, year = {1993}, month = {April}, volume = {8}, number = {4}, pages = {539-67}, abstract = {ABSTRACT Research on potential interactions between connectionist learning systems, i.e., artificial neural networks (ANNs), and evolutionary search procedures, like genetic algorithms (GAs), has attracted a lot of attention. Evolutionary ANNs (EANNs) can be considered as the combination of ANNs and evolutionary search procedures. This article first distinguishes among three kinds of evolution in EANNs, i.e., the evolution of connection weights, of architectures, and of learning rules. Then it reviews each kind of evolution in detail and analyzes critical issues related to different evolutions. The review shows that although a lot of work has been done on the evolution of connection weights and architectures, few attempts have been made to understand the evolution of learning rules. Interactions among different evolutions are seldom mentioned in current research. However, the evolution of learning rules and its interactions with other kinds of evolution, play a vital role in EANNs. Finally, this article briefly describes a general framework for EANNs, which not only includes the aforementioned three kinds of evolution, but also considers interactions among them.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } } @inproceedings{Zhang93, key = {connectionism, cogann}, author = {Byoung-Tak Zhang and H. M\H{u}hlenbein}, title = {Genetic Programming of Minimal Neural Nets Using Occam's Razor}, booktitle = {Proceedings of the International Conference on Genetic Algorithms}, year = {1993}, topology = { }, network = { }, encoding = {indirect, LISP program}, evolves = { }, applications = { } } @inproceedings{Zhang91, key = {genetic algorithms connectionism neural networks cogann}, author = {Byoung-Tak Zhang and Gerd Veenker}, title = {Neural Networks that Teach Themselves Through Genetic Discovery of Novel Examples}, booktitle = {Proceedings of the International Joint Conference on Neural Networks}, year = {1991}, pages = {690-695}, abstract = {ABSTRACT The Authors introduce "" active learning paradigm for neural networks. In contrast to the passive paradigm, the learning in the active paradigm is initiated by the machine learner instead of its environment or teacher. The Authors present " learning algorithm that uses a genetic algorithm for creating novel examples to teach multilayer feedforward networks. The creative learning networks, based on their own knowledge, discover new examples, criticize and select useful ones, train themselves, and thereby extend their existing knowledge. Experiments on function extrapolation show that the self-teaching neural networks not only reduce the teaching efforts of the human, but the genetically created examples also contribute robustly to the improvement of generalization performance and the interpretation of the connectionist knowledge.}, topology = { }, network = { }, encoding = { }, evolves = { }, applications = { } }