- author = "Yun Seog Yeun and K. H. Lee and Y. S. Yang",
- title = "Function approximations by coupling neural networks and genetic programming trees with oblique decision trees",
- journal = "Artificial Intelligence in Engineering",
- year = "1999",
- volume = "13",
- number = "3",
- pages = "223--239",
- month = jul,
- email = "yeonyun@road.daejin.ac.kr",
- keywords = "genetic algorithms, genetic programming, Federated agents, Oblique decision tree, OC1",
-
URL = "
http://members.kr.inter.net/yyshuj/paper/aie.zip",
- ISSN = "0954-1810",
-
DOI = "
10.1016/S0954-1810(98)00015-6",
-
URL = "http://www.sciencedirect.com/science/article/B6V1X-3WWT8F6-3/1/2e564ef70743de81b8e3369fb01b406e",
- abstract = "This paper is concerning the development of multiple neural networks system combined with genetic programming (GP) trees for problem domains where the complete input space can be decomposed into several different regions, and these are well represented in form of oblique decision tree. The overall architecture of hybrid system, called the federated agents, consists of a facilitator, local agents, and boundary agents. Neural networks used as local agents, each of which is expert at different subregions, and GP trees serve as boundary agents. A boundary agent refer to the one that is specialized at only the borders of subregions where discontinuities or different patterns may exist. The facilitator is responsible for choosing the local agent that is suitable for the given input data using information obtained from oblique decision tree representing a divided input space. However, there are large possibility of selecting the invalid local agent due to the incorrect prediction of decision tree, provided that input data is close enough to the boundaries of regions. Such a situation can lead federated agents to produce a much higher prediction error than that of a single neural network trained over all input space. To deal with this, the approach taken in this paper is to make the facilitator select the boundary agent instead of the local agent when input data is closely located to the certain border of regions. In this way, even if the result of decision tree may be incorrect, the results of system are less affected by it. The validity of our approach is examined and verified by applying the federated agents to the configuration design of a midship section of bulk cargo ships.",
- size = "17 pages",
-
notes = "Linear associative memories [Kohonen,1988] set
numerical parameters in GP trees with overfitting
avoidance.
Training set partitioned using {"}domain knowledge or clustering methods{"} p255. Separate ANN trained on each subset.
",
