DocumentCode :
2575768
Title :
Pathologies of temporal difference methods in approximate dynamic programming
Author :
Bertsekas, Dimitri P.
Author_Institution :
Dept. of Electr. Eng. & Comp. Sci., M.I.T., Cambridge, MA, USA
fYear :
2010
fDate :
15-17 Dec. 2010
Firstpage :
3034
Lastpage :
3039
Abstract :
Approximate policy iteration methods based on temporal differences are popular in practice, and have been tested extensively, dating to the early nineties, but the associated convergence behavior is complex, and not well understood at present. An important question is whether the policy iteration process is seriously hampered by oscillations between poor policies, roughly similar to the attraction of gradient methods to poor local minima. There has been little apparent concern in the approximate DP/reinforcement learning literature about this possibility, even though it has been documented with several simple examples. Recent computational experimentation with the game of tetris, a popular testbed for approximate DP algorithms over a 15-year period, has brought the issue to sharp focus. In particular, using a standard set of 22 features and temporal difference methods, an average score of a few thousands was achieved. Using the same features and a random search method, an overwhelmingly better average score was achieved (600,000-900,000). The paper explains the likely mechanism of this phenomenon, and derives conditions under which it will not occur.
Keywords :
dynamic programming; iterative methods; random processes; search problems; approximate dynamic programming; approximate policy iteration method; random search method; temporal difference method; Aggregates; Convergence; Equations; Function approximation; Mathematical model; Oscillators;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Decision and Control (CDC), 2010 49th IEEE Conference on
Conference_Location :
Atlanta, GA
ISSN :
0743-1546
Print_ISBN :
978-1-4244-7745-6
Type :
conf
DOI :
10.1109/CDC.2010.5717644
Filename :
5717644
Link To Document :
بازگشت