Complexity in Genetic Programming: Using Entropy and Compression Metrics to Understand GP Behaviour
Created by W.Langdon from
author = "R. I. (Bob) McKay",
howpublished = "International Symposium on Intelligence, Computation
address = "Wuhan, China",
month = sep,
note = "Keynote Address",
title = "Complexity in Genetic Programming: Using Entropy and
Compression Metrics to Understand GP Behaviour",
year = "2007",
keywords = "genetic algorithms, genetic programming, bloat, size
limit and introns, VC dimension",
URL = "http://sc.snu.ac.kr/PAPERS/entcompress.pdf",
notes = "'oppose convergence' eg 'Fitness sharing',
'Anticorrelation penalties', 'information based
accuracy, parsimony', 'Disappointing results so far',
'Regular structure virtually never emerges in GP
'Equivalent Decision Simplification'.
'Symbolic Regression' of cos(2x), '1000 runs, 500
population, 200 generations'.
'Information theory to understand genetic programming'.
'Entropy of (binary?) subtrees' up to 4
'Compression and Regularity Regularity'.
'compressibility to measure complexity'. 'XMLPPM, an
excellent tree compression algorithm'. 'Measure how
much different runs discover the same building
Genetic Programming entries for
R I (Bob) McKay