dc.contributor.author |
James, S
|
|
dc.contributor.author |
Konidaris, G
|
|
dc.contributor.author |
Rosman, Benjamin S
|
|
dc.date.accessioned |
2017-09-20T09:51:12Z |
|
dc.date.available |
2017-09-20T09:51:12Z |
|
dc.date.issued |
2017-02 |
|
dc.identifier.citation |
James, S., Konidaris, G., and Rosman, B.S. 2017. An analysis of Monte Carlo tree search. Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, San Francisco, California, USA, 4-9 February 2017 |
en_US |
dc.identifier.uri |
https://aaai.org/Library/AAAI/aaai17contents.php
|
|
dc.identifier.uri |
https://aaai.org/ocs/index.php/AAAI/AAAI17/paper/view/14886
|
|
dc.identifier.uri |
http://hdl.handle.net/10204/9582
|
|
dc.description |
Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, San Francisco, California, USA, 4-9 February 2017 |
en_US |
dc.description.abstract |
Monte Carlo Tree Search (MCTS) is a family of directed search algorithms that has gained widespread attention in recent years. Despite the vast amount of research into MCTS, the effect of modifications on the algorithm, as well as the manner in which it performs in various domains, is still not yet fully known. In particular, the effect of using knowledge heavy rollouts in MCTS still remains poorly understood, with surprising results demonstrating that better-informed rollouts often result in worse-performing agents. We present experimental evidence suggesting that, under certain smoothness conditions, uniformly random simulation policies preserve the ordering over action preferences. This explains the success of MCTS despite its common use of these rollouts to evaluate states. We further analyse non-uniformly random rollout policies and describe conditions under which they offer improved performance. |
en_US |
dc.language.iso |
en |
en_US |
dc.publisher |
Association for the Advancement of Artificial Intelligence |
en_US |
dc.relation.ispartofseries |
Worklist;19461 |
|
dc.subject |
Monte-Carlo Tree Search |
en_US |
dc.subject |
MCTS |
en_US |
dc.subject |
Roll-outs |
en_US |
dc.title |
An analysis of Monte Carlo tree search |
en_US |
dc.type |
Conference Presentation |
en_US |
dc.identifier.apacitation |
James, S., Konidaris, G., & Rosman, B. S. (2017). An analysis of Monte Carlo tree search. Association for the Advancement of Artificial Intelligence. http://hdl.handle.net/10204/9582 |
en_ZA |
dc.identifier.chicagocitation |
James, S, G Konidaris, and Benjamin S Rosman. "An analysis of Monte Carlo tree search." (2017): http://hdl.handle.net/10204/9582 |
en_ZA |
dc.identifier.vancouvercitation |
James S, Konidaris G, Rosman BS, An analysis of Monte Carlo tree search; Association for the Advancement of Artificial Intelligence; 2017. http://hdl.handle.net/10204/9582 . |
en_ZA |
dc.identifier.ris |
TY - Conference Presentation
AU - James, S
AU - Konidaris, G
AU - Rosman, Benjamin S
AB - Monte Carlo Tree Search (MCTS) is a family of directed search algorithms that has gained widespread attention in recent years. Despite the vast amount of research into MCTS, the effect of modifications on the algorithm, as well as the manner in which it performs in various domains, is still not yet fully known. In particular, the effect of using knowledge heavy rollouts in MCTS still remains poorly understood, with surprising results demonstrating that better-informed rollouts often result in worse-performing agents. We present experimental evidence suggesting that, under certain smoothness conditions, uniformly random simulation policies preserve the ordering over action preferences. This explains the success of MCTS despite its common use of these rollouts to evaluate states. We further analyse non-uniformly random rollout policies and describe conditions under which they offer improved performance.
DA - 2017-02
DB - ResearchSpace
DP - CSIR
KW - Monte-Carlo Tree Search
KW - MCTS
KW - Roll-outs
LK - https://researchspace.csir.co.za
PY - 2017
T1 - An analysis of Monte Carlo tree search
TI - An analysis of Monte Carlo tree search
UR - http://hdl.handle.net/10204/9582
ER -
|
en_ZA |