reference/modules_mcts

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

torchrl.modules.mcts package

This module provides Monte Carlo Tree Search (MCTS) components, including score computation modules for balancing exploration and exploitation in tree search algorithms.

MCTS Scores

MCTSScore(*args, **kwargs)	Abstract base class for MCTS score computation modules.
PUCTScore(*args, **kwargs)	Computes the PUCT (Polynomial Upper Confidence Trees) score for MCTS.
UCBScore(*args, **kwargs)	Computes the UCB (Upper Confidence Bound) score, specifically UCB1, for MCTS.
EXP3Score(*args, **kwargs)	Computes action selection probabilities for the EXP3 algorithm in MCTS.
UCB1TunedScore(*args, **kwargs)	Computes the UCB1-Tuned score for MCTS, using variance estimation.
MCTSScores(value[, names, module, qualname, ...])	A collection of MCTS score computation modules.