Co-evolving Fitness Predictors for Accelerating and
Reducing Evaluations
M. Schmidt, and H. Lipson. Genetic Programming Theory and Practice IV, volume 5 of Genetic and Evolutionary Computation, chapter 17, Springer, Ann Arbor, (11-13 May 2006)
Abstract
We investigate co-evolutionary GP that that co-evolves
fitness predictors in order to reduce the computational
cost of evolution and/or reduced the number of
evaluations required. Fitness predictors are light
objects which, given an evolving individual,
heuristically approximate its true fitness. The
predictors are trained by their ability to correctly
differentiate between good and bad solutions using
reduced computation. We apply coevolution of fitness
predictors to symbolic regression and measure its
impact. Our results show that a small computational
investment in co-evolving fitness predictors greatly
enhances both speed and convergence of individual
solutions while reducing the computational effort
overall. Finally we apply fitness prediction to
interactive evolution of pen stroke drawings. These
results show that fitness prediction is extremely
effective at modelling user preference while minimising
the sampling on the user to fewer than ten prompts.
%0 Book Section
%1 Schmidt:2006:GPTP
%A Schmidt, Michael D.
%A Lipson, Hod
%B Genetic Programming Theory and Practice IV
%C Ann Arbor
%D 2006
%E Riolo, Rick L.
%E Soule, Terence
%E Worzel, Bill
%I Springer
%K algorithms, genetic programming
%P -
%T Co-evolving Fitness Predictors for Accelerating and
Reducing Evaluations
%V 5
%X We investigate co-evolutionary GP that that co-evolves
fitness predictors in order to reduce the computational
cost of evolution and/or reduced the number of
evaluations required. Fitness predictors are light
objects which, given an evolving individual,
heuristically approximate its true fitness. The
predictors are trained by their ability to correctly
differentiate between good and bad solutions using
reduced computation. We apply coevolution of fitness
predictors to symbolic regression and measure its
impact. Our results show that a small computational
investment in co-evolving fitness predictors greatly
enhances both speed and convergence of individual
solutions while reducing the computational effort
overall. Finally we apply fitness prediction to
interactive evolution of pen stroke drawings. These
results show that fitness prediction is extremely
effective at modelling user preference while minimising
the sampling on the user to fewer than ten prompts.
%& 17
%@ 0-387-33375-4
@incollection{Schmidt:2006:GPTP,
abstract = {We investigate co-evolutionary GP that that co-evolves
fitness predictors in order to reduce the computational
cost of evolution and/or reduced the number of
evaluations required. Fitness predictors are light
objects which, given an evolving individual,
heuristically approximate its true fitness. The
predictors are trained by their ability to correctly
differentiate between good and bad solutions using
reduced computation. We apply coevolution of fitness
predictors to symbolic regression and measure its
impact. Our results show that a small computational
investment in co-evolving fitness predictors greatly
enhances both speed and convergence of individual
solutions while reducing the computational effort
overall. Finally we apply fitness prediction to
interactive evolution of pen stroke drawings. These
results show that fitness prediction is extremely
effective at modelling user preference while minimising
the sampling on the user to fewer than ten prompts.},
added-at = {2008-06-19T17:46:40.000+0200},
address = {Ann Arbor},
author = {Schmidt, Michael D. and Lipson, Hod},
biburl = {https://www.bibsonomy.org/bibtex/221860b99f7d4c3a7962756b120cee10d/brazovayeye},
booktitle = {Genetic Programming Theory and Practice {IV}},
chapter = 17,
editor = {Riolo, Rick L. and Soule, Terence and Worzel, Bill},
interhash = {50584e0a25c495fcc6e30cc7ae3066f9},
intrahash = {21860b99f7d4c3a7962756b120cee10d},
isbn = {0-387-33375-4},
keywords = {algorithms, genetic programming},
month = {11-13 May},
notes = {part of \cite{Riolo:2006:GPTP} Published Jan 2007
after the workshop},
pages = {-},
publisher = {Springer},
series = {Genetic and Evolutionary Computation},
timestamp = {2008-06-19T17:51:07.000+0200},
title = {Co-evolving Fitness Predictors for Accelerating and
Reducing Evaluations},
volume = 5,
year = 2006
}