Crate simple_mcts

Macros§

impl_resettable_buffer: Implements ResettableBuffer for standard collections.

Action: Represents an index corresponding to a specific legal move or action.
Engine: The core Engine handling the Monte Carlo Tree Search algorithm.
MctsAction: A structure bundling a game action with its targeted engine ID.
MctsBatch: The Batch Manager for MCTS Engines.
MctsConfig: Configuration for the MCTS batch manager.
MctsId: A unique, strongly-typed identifier representing a specific Engine instance within the MctsBatch manager.
MctsNodeId: A strictly typed identifier for an instantiated node within the MCTS tree.
MctsScore: A structure bundling the final score array with its targeted engine ID.
MctsSelectionResult: Represents the output of a selection phase for a specific engine.
MctsStateEvaluation: A structure bundling an MCTS evaluation with its targeted engine ID.

MctsEngineError: Errors that can occur during the execution of the MCTS engine.
MctsError: Errors that can occur when interacting with the MctsBatch manager.
SelectionResult: Represents the outcome of a tree selection phase.
StateEvaluation: Represents the evaluation of a game state, usually provided by a heuristic function or a neural network (e.g., in AlphaZero/MuZero architectures).

ResettableBuffer: A buffer used to store the sequence of actions traversed during the selection phase.
ScoreTransformer: Transforms a score during the backpropagation phase.
SelectionFunction: Evaluates a child node during the tree selection phase.

negate_score: Inverts the evaluation score during the backpropagation phase.
puct: Calculates the Predictor + Upper Confidence bound applied to Trees (PUCT) score.
visits_to_probabilities: Converts a raw array of visit counts into a normalized probability distribution.
visits_to_probabilities_with_temperature: Converts visit counts into a probability distribution using a temperature parameter.