Why Co-Evolution Beats Temporal Difference Learning at Backgammon for a Linear Architecture, but not a Non-Linear Architecture

by: Paul J Darwen

(FebruaryJuly-March0 2001), pp. 1003-1010.

View FullText article

CiteSeer (at Penn State), CiteSeer (at MIT), CiteSeer (at Zurich), CiteSeer (at Singapore)

Reviews [Write a review of this article]

There are no reviews of this article

Find related articles from these CiteULike users

pchiusano

Find related articles with these CiteULike tags

backgammon, coevolution, hill_climbing, td-learning

Abstract

This paper compares co-evolutionary learning and temporal difference learning on the game of Backgammon, which (like many real-world tasks) has an element of random uncertainty. Unfortunately, to fully evaluate a single strategy using undirected sampling of board positions, using only random dice rolls, requires a great deal of computation. Evolution's all-or-nothing replacement of entire solutions needs accurate evaluation, but relatively rare board positions are needed to train above a...

@inproceedings{citeulike:515122, abstract = {This paper compares co-evolutionary learning and temporal difference learning on the game of Backgammon, which (like many real-world tasks) has an element of random uncertainty. Unfortunately, to fully evaluate a single strategy using undirected sampling of board positions, using only random dice rolls, requires a great deal of computation. Evolution's all-or-nothing replacement of entire solutions needs accurate evaluation, but relatively rare board positions are needed to train above a...}, address = {COEX, World Trade Center, 159 Samseong-dong, Gangnam-gu, Seoul, Korea}, author = {Darwen, Paul J. }, booktitle = {Proceedings of the 2001 Congress on Evolutionary Computation CEC2001}, citeulike-article-id = {515122}, keywords = {backgammon, coevolution, hill\_climbing, td-learning}, month = {FebruaryJuly-March0}, pages = {1003--1010}, posted-at = {2006-02-21 19:12:02}, priority = {2}, publisher = {IEEE Press}, title = {Why Co-Evolution Beats Temporal Difference Learning at Backgammon for a Linear Architecture, but not a Non-Linear Architecture}, url = {http://citeseer.ist.psu.edu/445415.html}, year = {2001} }

RIS record

TY - CONF ID - citeulike:515122 L3 - citeulike-article-id:515122 TI - Why Co-Evolution Beats Temporal Difference Learning at Backgammon for a Linear Architecture, but not a Non-Linear Architecture T2 - Proceedings of the 2001 Congress on Evolutionary Computation CEC2001 SP - 1003 EP - 1010 AD - COEX, World Trade Center, 159 Samseong-dong, Gangnam-gu, Seoul, Korea PB - IEEE Press N2 - This paper compares co-evolutionary learning and temporal difference learning on the game of Backgammon, which (like many real-world tasks) has an element of random uncertainty. Unfortunately, to fully evaluate a single strategy using undirected sampling of board positions, using only random dice rolls, requires a great deal of computation. Evolution's all-or-nothing replacement of entire solutions needs accurate evaluation, but relatively rare board positions are needed to train above a... KW - backgammon KW - coevolution KW - hill_climbing KW - td-learning AU - Darwen, Paul PY - 2001/27-30// UR - http://citeseer.ist.psu.edu/445415.html ER -

RIS BibTeX