Implementing temporal-difference learning with the scaled conjugate gradient algorithm

Falas, T; Stafylopatis, A

dc.contributor.author	Falas, T	en
dc.contributor.author	Stafylopatis, A	en
dc.date.accessioned	2014-03-01T01:22:30Z
dc.date.available	2014-03-01T01:22:30Z
dc.date.issued	2005	en
dc.identifier.issn	1370-4621	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/16594
dc.subject	Reinforcement learning	en
dc.subject	Scaled conjugate gradient	en
dc.subject	Temporal-difference learning	en
dc.subject	Time series prediction	en
dc.subject.classification	Computer Science, Artificial Intelligence	en
dc.subject.classification	Neurosciences	en
dc.subject.other	Data processing	en
dc.subject.other	Learning algorithms	en
dc.subject.other	Linear systems	en
dc.subject.other	Problem solving	en
dc.subject.other	Time series analysis	en
dc.subject.other	Reinforcement learning	en
dc.subject.other	Scaled conjugate gradients	en
dc.subject.other	Temporal-difference learning	en
dc.subject.other	Time series prediction	en
dc.subject.other	Learning systems	en
dc.title	Implementing temporal-difference learning with the scaled conjugate gradient algorithm	en
heal.type	journalArticle	en
heal.identifier.primary	10.1007/s11063-005-1384-x	en
heal.identifier.secondary	http://dx.doi.org/10.1007/s11063-005-1384-x	en
heal.language	English	en
heal.publicationDate	2005	en
heal.abstract	This paper investigates the use of the scaled conjugate gradient (SCG) algorithm in temporal-difference (TD) learning for time series prediction. Special emphasis is given on the implementation details, after examining the theoretical background of the algorithm and the learning methodology and how these could be combined. Simple time series (linear, sinusoidal, etc.) as well as more complex ones, coming from real data, are used to examine the behavior of this novel combination of learning algorithm and methodology. Preliminary experimental results indicate that the implementation as presented in this paper indeed works, but the performance (in terms of learning speed and generalization ability) of TD learning using the SCG algorithm is not as good as expected, at least on the representative problems examined. An attempt to rationalize these results is presented. © Springer 2005.	en
heal.publisher	SPRINGER	en
heal.journalName	Neural Processing Letters	en
dc.identifier.doi	10.1007/s11063-005-1384-x	en
dc.identifier.isi	ISI:000233276900009	en
dc.identifier.volume	22	en
dc.identifier.issue	3	en
dc.identifier.spage	361	en
dc.identifier.epage	375	en