Real-Time Search Method in Nondeterministic Game — Ms. Pac-Man

Gan, Xiaocong; Bao, Yun; Han, Zhangang

doi:10.3233/icg-2011-34404

Cited by 3 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MCTS has had success in problems with high branching factor, and its benefit is best realized when adapted to suit the target domain (Chaslot et al, ). This includes Atari games such as Ms Pac‐Man (Gan, Bao, & Han, ), and Total War: Rome II (Champandard, ). The most notable accomplishment of MCTS is its combination with deep neural networks to beat one of the best human players (Silver et al, ).…”

Section: Introductionmentioning

confidence: 99%

Learning decision trees through Monte Carlo tree search: An empirical evaluation

Nunes

Craene

Langet

et al. 2020

WIREs Data Min & Knowl

View full text Add to dashboard Cite

Decision trees (DTs) are a widely used prediction tool, owing to their interpretability. Standard learning methods follow a locally optimal approach that trades off prediction performance for computational efficiency. Such methods can however be far from optimal, and it may pay off to spend more computational resources to increase performance. Monte Carlo tree search (MCTS) is an approach to approximate optimal choices in exponentially large search spaces. We propose a DT learning approach based on the Upper Confidence Bound applied to tree (UCT) algorithm, including procedures to expand and explore the space of DTs. To mitigate the computational cost of our method, we employ search pruning strategies that discard some branches of the search tree. The experiments show that proposed approach outperformed the C4.5 algorithm in 20 out of 31 datasets, with statistically significant improvements in the trade-off between prediction performance and DT complexity. The approach improved locally optimal search for datasets with more than 1,000 instances, or for smaller datasets likely arising from complex distributions.

show abstract

Section: Introductionmentioning

confidence: 99%

Learning decision trees through Monte Carlo tree search: An empirical evaluation

Nunes

Craene

Langet

et al. 2020

WIREs Data Min & Knowl

View full text Add to dashboard Cite

show abstract

“…The MCTS algorithm has had success in problems with high branching factor, and its benefit is best realized when adapted to suit the domain at hand [10]. This includes playing Atari games such as Ms Pac-Man [11], and Total War: Rome II [12], and computer Go [13].…”

Section: Introductionmentioning

confidence: 99%

A Monte Carlo Tree Search Approach to Learning Decision Trees

Nunes

Craene

Langet

et al. 2018

2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA)

View full text Add to dashboard Cite

Decision trees (DTs) are a widely used prediction tool, owing to their interpretability. Standard learning methods follow a locally-optimal approach that trades off prediction performance for computational efficiency. Such methods can however be far from optimal, and it may pay off to spend more computational resources to increase performance. Monte Carlo tree search (MCTS) is an approach to approximate optimal choices in exponentially large search spaces. Since exploring the space of all possible DTs is computationally intractable, we propose a DT learning approach based on MCTS. To bound the branching factor of MCTS, we limit the number of decisions at each level of the search tree, and introduce mechanisms to balance exploration, DT size and the statistical significance of the predictions. To mitigate the computational cost of our method, we employ a move pruning strategy that discards some branches of the search tree, leading to improved performance. The experiments show that our approach outperformed locallyoptimal search in 20 out of 31 datasets, with a reduction in DT size in most of the cases.

show abstract

A Rule-Based AI Method for an Agent Playing Big Two

2021

View full text Add to dashboard Cite

Big Two is a popular multiplayer card game in Asia. This research proposes a new method, named rule-based AI, for an agent playing Big Two. The novel method derives rules based on the number of cards the AI agent has left and the control position. The rules for two to four cards left are used to select the card combination to discard based on the number of cards remaining in the agent’s hand. The rules for more than four cards left conditionally prioritize discarding the card combination in the classified cards with lower priority. A winning strategy provides guidelines to guarantee that the AI agent will win when a win is achievable within three moves. We also design the rules for the AI agent without control for holding cards and splitting cards. The experimental results show that our proposed AI agent can play Big Two well and outperform randomized AI, conventional AI, and human players, presenting winning rates of 89.60%, 73.00%, and 55.05%, respectively, with the capability of maximizing the winning score and minimizing the number of cards left when the chance of winning is low.

show abstract

Real-Time Search Method in Nondeterministic Game — Ms. Pac-Man

Cited by 3 publications

References 0 publications

Learning decision trees through Monte Carlo tree search: An empirical evaluation

Learning decision trees through Monte Carlo tree search: An empirical evaluation

A Monte Carlo Tree Search Approach to Learning Decision Trees

A Rule-Based AI Method for an Agent Playing Big Two

Contact Info

Product

Resources

About