dc.contributor.author |
James, S
|
|
dc.contributor.author |
Rosman, Benjamin S
|
|
dc.contributor.author |
Konidaris, G
|
|
dc.date.accessioned |
2017-02-23T09:59:05Z |
|
dc.date.available |
2017-02-23T09:59:05Z |
|
dc.date.issued |
2016-07 |
|
dc.identifier.citation |
James, S., Rosman, B.S. and Konidaris, G. 2016. An investigation into the effectiveness of heavy rollouts in UCT. In: General Intelligence in Game-Playing Agents (GIGA'16) Workshop at IJCAI'16, 9-15 July 2016, New York City, New York |
en_US |
dc.identifier.uri |
https://www.benjaminrosman.com/papers/ijcai16.pdf
|
|
dc.identifier.uri |
http://hdl.handle.net/10204/8942
|
|
dc.description |
General Intelligence in Game-Playing Agents (GIGA'16) Workshop at IJCAI'16, 9-15 July 2016, New York City, New York |
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, with its domain-independent nature making it particularly attractive to fields such as General Game Playing. Despite the vast amount of research into MCTS, the dynamics of the algorithm are still not yet fully understood. In particular, the effect of using knowledge-heavy or biased rollouts in MCTS still remains largely unknown, with surprising results demonstrating that better-informed rollouts do not necessarily result in stronger agents. We show that MCTS is well-suited to a class of domains possessing a smoothness property, and that any error due to incorrect bias is compounded in non-smooth domains, particularly for low-variance simulations. |
en_US |
dc.language.iso |
en |
en_US |
dc.relation.ispartofseries |
Wokflow;17893 |
|
dc.subject |
Monte Carlo tree search |
en_US |
dc.subject |
Biased rollouts |
en_US |
dc.subject |
MCTS |
en_US |
dc.title |
An investigation into the effectiveness of heavy rollouts in UCT |
en_US |
dc.type |
Conference Presentation |
en_US |
dc.identifier.apacitation |
James, S., Rosman, B. S., & Konidaris, G. (2016). An investigation into the effectiveness of heavy rollouts in UCT. http://hdl.handle.net/10204/8942 |
en_ZA |
dc.identifier.chicagocitation |
James, S, Benjamin S Rosman, and G Konidaris. "An investigation into the effectiveness of heavy rollouts in UCT." (2016): http://hdl.handle.net/10204/8942 |
en_ZA |
dc.identifier.vancouvercitation |
James S, Rosman BS, Konidaris G, An investigation into the effectiveness of heavy rollouts in UCT; 2016. http://hdl.handle.net/10204/8942 . |
en_ZA |
dc.identifier.ris |
TY - Conference Presentation
AU - James, S
AU - Rosman, Benjamin S
AU - Konidaris, G
AB - Monte Carlo Tree Search (MCTS) is a family of directed search algorithms that has gained widespread attention in recent years, with its domain-independent nature making it particularly attractive to fields such as General Game Playing. Despite the vast amount of research into MCTS, the dynamics of the algorithm are still not yet fully understood. In particular, the effect of using knowledge-heavy or biased rollouts in MCTS still remains largely unknown, with surprising results demonstrating that better-informed rollouts do not necessarily result in stronger agents. We show that MCTS is well-suited to a class of domains possessing a smoothness property, and that any error due to incorrect bias is compounded in non-smooth domains, particularly for low-variance simulations.
DA - 2016-07
DB - ResearchSpace
DP - CSIR
KW - Monte Carlo tree search
KW - Biased rollouts
KW - MCTS
LK - https://researchspace.csir.co.za
PY - 2016
T1 - An investigation into the effectiveness of heavy rollouts in UCT
TI - An investigation into the effectiveness of heavy rollouts in UCT
UR - http://hdl.handle.net/10204/8942
ER -
|
en_ZA |