- author = "Krzysztof Krawiec and Armando Solar-Lezama",
- title = "Improving Genetic Programming with Behavioral Consistency Measure",
- booktitle = "13th International Conference on Parallel Problem Solving from Nature",
- year = "2014",
- editor = "Thomas Bartz-Beielstein and Juergen Branke and Bogdan Filipic and Jim Smith",
- publisher = "Springer",
- isbn13 = "978-3-319-10761-5",
- pages = "434--443",
- series = "Lecture Notes in Computer Science",
- address = "Ljubljana, Slovenia",
- month = "13-17 " # sep,
- volume = "8672",
- keywords = "genetic algorithms, genetic programming, Program synthesis, entropy, information theory, information consistency, multiobjective search",
-
DOI = "
doi:10.1007/978-3-319-10762-2_43",
- abstract = "Program synthesis tasks usually specify only the desired output of a program and do not state any expectations about its internal behaviour. The intermediate execution states reached by a running program can be nonetheless deemed as more or less preferred according to their information content with respect to the desired output. A consistency measure is proposed that implements this observation. When used as an additional search objective in a typical genetic programming setting, this measure improves the success rate on a suite of 35 benchmarks in a statistically significant way.",
-
notes = "
Tree based GP, NSGA-II, ECJ, 250 generations, pop=1000 tournament size 5, no constants.
log_2(3) = 1.58496
Fitness how close to being correct from internal intermediate values of the program. Ie do intermediate values preserve important information about the inputs?
Add break points (in fixed order, ie outside ifs and loops) and examine memory's contents at them (in GP root of subtrees).
