- author = "Walter Alden Tackett",
- title = "Greedy Recombination and Genetic Search on the Space of Computer Programs",
- booktitle = "Foundations of Genetic Algorithms 3",
- year = "1994",
- editor = "L. Darrell Whitley and Michael D. Vose",
- pages = "271--297",
- publisher_address = "San Francisco, CA, USA",
- address = "Estes Park, Colorado, USA",
- month = "31 " # jul # "--2 " # aug,
- organisation = "International Society for Genetic Algorithms",
- publisher = "Morgan Kaufmann",
- note = "Published 1995",
- keywords = "genetic algorithms, genetic programming",
- ISBN = "1-55860-356-5",
-
URL = "
http://www.amazon.co.uk/gp/search?index=books&linkCode=qs&keywords=1558603565",
-
DOI = "
doi:10.1016/B978-1-55860-356-1.50017-0",
- abstract = "Many natural organisms overproduce zygotes and subsequently decimate the ranks of offspring at some later stage of development. The basic purpose of this behavior is the reduction of parental resource investment in offspring which are less fit than others according to some metabolically cheap fitness measure. An important insight into this process is that for single-pair matings all offspring are products of the same parental genotypes: the selection taking place therefore seeks the most fit recombination of parental traits. This paper presents the Greedy Recombination operator RB(n) for genetic programming, which performs greedy selection among potential crossover sites of a mating pair. The properties of RB(n) are described both from a statistical standpoint and in terms of their effect upon search; comparisons are drawn to existing methods. We formulate the class of constructional problems, which allow precise control over the fitness structure in the space of expressions being searched. The constructional approach is used to create simple GP analogies of the Royal Road problems used to study classical GA. The effects of RB(n) upon search properties, fitness distributions, and genotypic variations are examined and contrasted with effects of selection and recombination methods.",
-
notes = "FOGA-3
Royal Road, Deceptive Royal Road, Soft brood selection, comparision with evolution strategies (u,l)-ES, NOP_1, NOP_2, NOP_4 (intron like). {"}Performance (of SSGA) depends critically on how individuals are selected for replacement (ref De Jong FOGA-2){"} page 288. {"}greatest fitness variation is always achieved using the tournament selection method{"} p290. Claims to explain results in \cite{kinnear:kinnear}
",
