Saved in:
Bibliographic Details
Main Authors: Kwak, Yunhyeok, Hwang, Inwoo, Kim, Dooyoung, Lee, Sanghack, Zhang, Byoung-Tak
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.00614
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914819321364480
author Kwak, Yunhyeok
Hwang, Inwoo
Kim, Dooyoung
Lee, Sanghack
Zhang, Byoung-Tak
author_facet Kwak, Yunhyeok
Hwang, Inwoo
Kim, Dooyoung
Lee, Sanghack
Zhang, Byoung-Tak
contents Monte Carlo Tree Search (MCTS) has showcased its efficacy across a broad spectrum of decision-making problems. However, its performance often degrades under vast combinatorial action space, especially where an action is composed of multiple sub-actions. In this work, we propose an action abstraction based on the compositional structure between a state and sub-actions for improving the efficiency of MCTS under a factored action space. Our method learns a latent dynamics model with an auxiliary network that captures sub-actions relevant to the transition on the current state, which we call state-conditioned action abstraction. Notably, it infers such compositional relationships from high-dimensional observations without the known environment model. During the tree traversal, our method constructs the state-conditioned action abstraction for each node on-the-fly, reducing the search space by discarding the exploration of redundant sub-actions. Experimental results demonstrate the superior sample efficiency of our method compared to vanilla MuZero, which suffers from expansive action space.
format Preprint
id arxiv_https___arxiv_org_abs_2406_00614
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Efficient Monte Carlo Tree Search via On-the-Fly State-Conditioned Action Abstraction
Kwak, Yunhyeok
Hwang, Inwoo
Kim, Dooyoung
Lee, Sanghack
Zhang, Byoung-Tak
Machine Learning
Artificial Intelligence
Monte Carlo Tree Search (MCTS) has showcased its efficacy across a broad spectrum of decision-making problems. However, its performance often degrades under vast combinatorial action space, especially where an action is composed of multiple sub-actions. In this work, we propose an action abstraction based on the compositional structure between a state and sub-actions for improving the efficiency of MCTS under a factored action space. Our method learns a latent dynamics model with an auxiliary network that captures sub-actions relevant to the transition on the current state, which we call state-conditioned action abstraction. Notably, it infers such compositional relationships from high-dimensional observations without the known environment model. During the tree traversal, our method constructs the state-conditioned action abstraction for each node on-the-fly, reducing the search space by discarding the exploration of redundant sub-actions. Experimental results demonstrate the superior sample efficiency of our method compared to vanilla MuZero, which suffers from expansive action space.
title Efficient Monte Carlo Tree Search via On-the-Fly State-Conditioned Action Abstraction
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2406.00614