Saved in:
Bibliographic Details
Main Authors: Dao, Mawaba Pascal, Peter, Adrian M.
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2501.13083
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909465060573184
author Dao, Mawaba Pascal
Peter, Adrian M.
author_facet Dao, Mawaba Pascal
Peter, Adrian M.
contents Active Inference, grounded in the Free Energy Principle, provides a powerful lens for understanding how agents balance exploration and goal-directed behavior in uncertain environments. Here, we propose a new planning framework, that integrates Monte Carlo Tree Search (MCTS) with active inference objectives to systematically reduce epistemic uncertainty while pursuing extrinsic rewards. Our key insight is that MCTS already renowned for its search efficiency can be naturally extended to incorporate free energy minimization by blending expected rewards with information gain. Concretely, the Cross-Entropy Method (CEM) is used to optimize action proposals at the root node, while tree expansions leverage reward modeling alongside intrinsic exploration bonuses. This synergy allows our planner to maintain coherent estimates of value and uncertainty throughout planning, without sacrificing computational tractability. Empirically, we benchmark our planner on a diverse set of continuous control tasks, where it demonstrates performance gains over both standalone CEM and MCTS with random rollouts.
format Preprint
id arxiv_https___arxiv_org_abs_2501_13083
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Boosting MCTS with Free Energy Minimization
Dao, Mawaba Pascal
Peter, Adrian M.
Artificial Intelligence
Active Inference, grounded in the Free Energy Principle, provides a powerful lens for understanding how agents balance exploration and goal-directed behavior in uncertain environments. Here, we propose a new planning framework, that integrates Monte Carlo Tree Search (MCTS) with active inference objectives to systematically reduce epistemic uncertainty while pursuing extrinsic rewards. Our key insight is that MCTS already renowned for its search efficiency can be naturally extended to incorporate free energy minimization by blending expected rewards with information gain. Concretely, the Cross-Entropy Method (CEM) is used to optimize action proposals at the root node, while tree expansions leverage reward modeling alongside intrinsic exploration bonuses. This synergy allows our planner to maintain coherent estimates of value and uncertainty throughout planning, without sacrificing computational tractability. Empirically, we benchmark our planner on a diverse set of continuous control tasks, where it demonstrates performance gains over both standalone CEM and MCTS with random rollouts.
title Boosting MCTS with Free Energy Minimization
topic Artificial Intelligence
url https://arxiv.org/abs/2501.13083