|
Title:
|
EXPLORING TO LEARN WINNING STRATEGY |
|
Author(s):
|
Samuel Lee |
|
ISBN:
|
978-989-8533-91-3 |
|
Editors:
|
Katherine Blashki and Yingcai Xiao |
|
Year:
|
2019 |
|
Edition:
|
Single |
|
Keywords:
|
Board Game, MCTS, Game Strategy, A.I. |
|
Type:
|
Short Paper |
|
First Page:
|
377 |
|
Last Page:
|
380 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
This paper presents a novel algorithm UCB* (called UCB star) to make intelligent decisions, enhancing the
winning chances within games through the utilization of a well-known tree search algorithm, Monte Carlo Tree Search
(MCTS). In the typical setting of a MCTS, exploration and exploitation are well balanced to propose optimal
decisions, given the game status of a board. This happens through the use of the Upper Confidence Bound (UCB).
In this paper, a new form of UCB is formulated by taking into account the posterior of the parameters, allowing bias
to be minimized from the observations.
Based on the experiments this paper performed, our UCB* has shown statistically significance of its effectiveness,
compared to a traditional UCB algorithm within the MCTS. |
|
|
|
|
|
|
