Evolving problem heuristics with on-line ACGP

author = "Cezary Z. Janikow",

title = "Evolving problem heuristics with on-line {ACGP}",

booktitle = "Late breaking paper at Genetic and Evolutionary Computation Conference {(GECCO'2007)}",

year = "2007",

month = "7-11 " # jul,

editor = "Peter A. N. Bosman",

isbn13 = "978-1-59593-698-1",

pages = "2503--2508",

address = "London, United Kingdom",

keywords = "genetic algorithms, genetic programming, heuristics, machine learning, STGP, artificial ant",

URL = "

http://gpbib.cs.ucl.ac.uk/gecco2007/docs/p2503.pdf",

DOI = "

doi:10.1145/1274000.1274017",

publisher = "ACM Press",

publisher_address = "New York, NY, USA",

size = "6 pages",

abstract = "Genetic Programming uses trees to represent chromosomes. The user defines the representation space by defining the set of functions and terminals to label the nodes in the trees. The sufficiency principle requires that the set be sufficient to label the desired solution trees, often forcing the user to enlarge the set, thus also enlarging the search space. Structure-preserving crossover, STGP, CGP, and CFG-based GP give the user the power to reduce the space by specifying rules for valid tree construction: types, syntax, and heuristics. However, in general the user may not be aware of the best representation space, including heuristics, to solve a particular problem. Recently, the ACGP methodology for extracting problem-specific heuristics, and thus for learning model of the problem domain, was introduced with preliminary off-line results. This paper overviews ACGP, pointing out its strength and limitations in the off-line mode. It then introduces a new on-line model, for learning while solving a problem, illustrated with experiments involving the multiplexer and the Santa Fe trail.",

notes = "Distributed on CD-ROM at GECCO-2007 ACM Order No. 910071",