Addressing the Even-n-parity problem using Compressed
Linear Genetic Programming
J. Parent, A. Nowe, and A. Defaweux. Late breaking paper at Genetic and Evolutionary
Computation Conference (GECCO'2005), Washington, D.C., USA, (25-29 June 2005)
Abstract
Compressed Linear Genetic Programming (cl-GP) uses
substring compression as a modularisation scheme.
Despite the fact that the compression of substrings
assumes a tight linkage between alleles, this approach
improves the GP search process. The compression of the
genotype, which is a form of linkage learning, provides
both a protection mechanism and a form of genetic code
reuse. This text presents the results obtained with the
cl-GP on the Even-n-parity problem. Results indicate
that the modularization of the cl-GP performs better
than a normal l-GP as it allows the cl-GP to preserve
useful gene combinations. Additionally the cl-GP
modularisation is well suited for problems where the
problem size is adjusted in a co-evolutionary setup,
the problem size increases each time a solution is
found
Late breaking paper at Genetic and Evolutionary
Computation Conference (GECCO'2005)
year
2005
month
25-29 June
notes
Distributed on CD-ROM at GECCO-2005
Pairs of adjacent functions and/or terminals present in
large numbers in 10 fit programs may be replaced by a
single symbol before crossover and mutation. The
intention being to keep them together as a building
block.
Representation is a linearised (depth first) tree. Non
standard meaning given to "co-evolutionary".
Up to even-10-parity evolved (cf poli:1999:22par
22 parity). Tight limit on program size. NOOP.
Elitism.
Why does size of dictionary rise after generation
zero?
%0 Conference Paper
%1 Parent:gecco05lbp
%A Parent, Johan
%A Nowe, Annie
%A Defaweux, Anne
%B Late breaking paper at Genetic and Evolutionary
Computation Conference (GECCO'2005)
%C Washington, D.C., USA
%D 2005
%E Rothlauf, Franz
%K algorithms, blocks building genetic modularisation, modules, programming,
%T Addressing the Even-n-parity problem using Compressed
Linear Genetic Programming
%U http://www.cs.bham.ac.uk/~wbl/biblio/gecco2005lbp/papers/54-parent.pdf
%X Compressed Linear Genetic Programming (cl-GP) uses
substring compression as a modularisation scheme.
Despite the fact that the compression of substrings
assumes a tight linkage between alleles, this approach
improves the GP search process. The compression of the
genotype, which is a form of linkage learning, provides
both a protection mechanism and a form of genetic code
reuse. This text presents the results obtained with the
cl-GP on the Even-n-parity problem. Results indicate
that the modularization of the cl-GP performs better
than a normal l-GP as it allows the cl-GP to preserve
useful gene combinations. Additionally the cl-GP
modularisation is well suited for problems where the
problem size is adjusted in a co-evolutionary setup,
the problem size increases each time a solution is
found
@inproceedings{Parent:gecco05lbp,
abstract = {Compressed Linear Genetic Programming (cl-GP) uses
substring compression as a modularisation scheme.
Despite the fact that the compression of substrings
assumes a tight linkage between alleles, this approach
improves the GP search process. The compression of the
genotype, which is a form of linkage learning, provides
both a protection mechanism and a form of genetic code
reuse. This text presents the results obtained with the
cl-GP on the Even-n-parity problem. Results indicate
that the modularization of the cl-GP performs better
than a normal l-GP as it allows the cl-GP to preserve
useful gene combinations. Additionally the cl-GP
modularisation is well suited for problems where the
problem size is adjusted in a co-evolutionary setup,
the problem size increases each time a solution is
found},
added-at = {2008-06-19T17:46:40.000+0200},
address = {Washington, D.C., USA},
author = {Parent, Johan and Nowe, Annie and Defaweux, Anne},
biburl = {https://www.bibsonomy.org/bibtex/2ee2f38bbef4b0218e967cb7dcf1e9add/brazovayeye},
booktitle = {Late breaking paper at Genetic and Evolutionary
Computation Conference {(GECCO'2005)}},
editor = {Rothlauf, Franz},
interhash = {ef7fced98021ad48dc4ff4e866298751},
intrahash = {ee2f38bbef4b0218e967cb7dcf1e9add},
keywords = {algorithms, blocks building genetic modularisation, modules, programming,},
month = {25-29 June},
notes = {Distributed on CD-ROM at GECCO-2005
Pairs of adjacent functions and/or terminals present in
large numbers in 10 fit programs may be replaced by a
single symbol before crossover and mutation. The
intention being to keep them together as a building
block.
Representation is a linearised (depth first) tree. Non
standard meaning given to {"}co-evolutionary{"}.
Up to even-10-parity evolved (cf \cite{poli:1999:22par}
22 parity). Tight limit on program size. NOOP.
Elitism.
Why does size of dictionary rise after generation
zero?},
timestamp = {2008-06-19T17:49:14.000+0200},
title = {Addressing the Even-n-parity problem using Compressed
Linear Genetic Programming},
url = {http://www.cs.bham.ac.uk/~wbl/biblio/gecco2005lbp/papers/54-parent.pdf},
year = 2005
}