Faster Genetic Programming based on Local Gradient
Search of Numeric Leaf Values
A. Topchy, and W. Punch. Proceedings of the Genetic and Evolutionary
Computation Conference (GECCO-2001), page 155--162. San Francisco, California, USA, Morgan Kaufmann, (7-11 July 2001)
Abstract
We examine the effectiveness of gradient search
optimization of numeric leaf values for Genetic
Programming. Genetic search for tree-like programs at
the population level is complemented by the
optimization of terminal values at the individual
level. Local adaptation of individuals is made easier
by algorithmic differentiation. We show how
conventional random constants are tuned by gradient
descent with minimal overhead. Several experiments with
symbolic regression problems are performed to
demonstrate the approach's effectiveness. Effects of
local learning are clearly manifest in both improved
approximation accuracy and selection changes when
periods of local and global search are interleaved.
Special attention is paid to the low overhead of the
local gradient descent. Finally, the inductive bias of
local learning is quantified.
Proceedings of the Genetic and Evolutionary
Computation Conference (GECCO-2001)
year
2001
month
7-11 July
pages
155--162
publisher
Morgan Kaufmann
publisher_address
San Francisco, CA 94104, USA
isbn
1-55860-774-9
notes
GECCO-2001 A joint meeting of the tenth International
Conference on Genetic Algorithms (ICGA-2001) and the
sixth Annual Genetic Programming Conference (GP-2001)
Part of spector:2001:GECCO
%0 Conference Paper
%1 topchy:2001:gecco
%A Topchy, Alexander
%A Punch, William F.
%B Proceedings of the Genetic and Evolutionary
Computation Conference (GECCO-2001)
%C San Francisco, California, USA
%D 2001
%E Spector, Lee
%E Goodman, Erik D.
%E Wu, Annie
%E Langdon, W. B.
%E Voigt, Hans-Michael
%E Gen, Mitsuo
%E Sen, Sandip
%E Dorigo, Marco
%E Pezeshk, Shahram
%E Garzon, Max H.
%E Burke, Edmund
%I Morgan Kaufmann
%K Baldwin Lamarckian algorithmic, algorithms, differentiation, effect, genetic gradient learning, optimization, programming, regression symbolic
%P 155--162
%T Faster Genetic Programming based on Local Gradient
Search of Numeric Leaf Values
%U http://www.cs.bham.ac.uk/~wbl/biblio/gecco2001/d01.pdf
%X We examine the effectiveness of gradient search
optimization of numeric leaf values for Genetic
Programming. Genetic search for tree-like programs at
the population level is complemented by the
optimization of terminal values at the individual
level. Local adaptation of individuals is made easier
by algorithmic differentiation. We show how
conventional random constants are tuned by gradient
descent with minimal overhead. Several experiments with
symbolic regression problems are performed to
demonstrate the approach's effectiveness. Effects of
local learning are clearly manifest in both improved
approximation accuracy and selection changes when
periods of local and global search are interleaved.
Special attention is paid to the low overhead of the
local gradient descent. Finally, the inductive bias of
local learning is quantified.
%@ 1-55860-774-9
@inproceedings{topchy:2001:gecco,
abstract = {We examine the effectiveness of gradient search
optimization of numeric leaf values for Genetic
Programming. Genetic search for tree-like programs at
the population level is complemented by the
optimization of terminal values at the individual
level. Local adaptation of individuals is made easier
by algorithmic differentiation. We show how
conventional random constants are tuned by gradient
descent with minimal overhead. Several experiments with
symbolic regression problems are performed to
demonstrate the approach's effectiveness. Effects of
local learning are clearly manifest in both improved
approximation accuracy and selection changes when
periods of local and global search are interleaved.
Special attention is paid to the low overhead of the
local gradient descent. Finally, the inductive bias of
local learning is quantified.},
added-at = {2008-06-19T17:46:40.000+0200},
address = {San Francisco, California, USA},
author = {Topchy, Alexander and Punch, William F.},
biburl = {https://www.bibsonomy.org/bibtex/2a55fc560e6774397b5df2468379c5cb2/brazovayeye},
booktitle = {Proceedings of the Genetic and Evolutionary
Computation Conference (GECCO-2001)},
editor = {Spector, Lee and Goodman, Erik D. and Wu, Annie and Langdon, W. B. and Voigt, Hans-Michael and Gen, Mitsuo and Sen, Sandip and Dorigo, Marco and Pezeshk, Shahram and Garzon, Max H. and Burke, Edmund},
interhash = {1db235ce2dd6240490a1ad2aed057185},
intrahash = {a55fc560e6774397b5df2468379c5cb2},
isbn = {1-55860-774-9},
keywords = {Baldwin Lamarckian algorithmic, algorithms, differentiation, effect, genetic gradient learning, optimization, programming, regression symbolic},
month = {7-11 July},
notes = {GECCO-2001 A joint meeting of the tenth International
Conference on Genetic Algorithms (ICGA-2001) and the
sixth Annual Genetic Programming Conference (GP-2001)
Part of \cite{spector:2001:GECCO}},
pages = {155--162},
publisher = {Morgan Kaufmann},
publisher_address = {San Francisco, CA 94104, USA},
timestamp = {2008-06-19T17:53:15.000+0200},
title = {Faster Genetic Programming based on Local Gradient
Search of Numeric Leaf Values},
url = {http://www.cs.bham.ac.uk/~wbl/biblio/gecco2001/d01.pdf},
year = 2001
}