Enhancements

Home

About

Tutorial

Applications

Enhancements

Bibliography

Resources

Project

Most MCTS research is devoted to improving the efficiency and resulting playing strength of the algorithm through various enhancements. We list some of the key enhancements below.


Bandit Based Methods

Bandit based methods lie at the heart of MCTS node selection. Most implementations use an Upper Confidence Bounds (UCB) approach or one of its derivatives.

UCB1
Text...

See: Ref (2006).

UCB1-Tuned
Text...

See: Ref (2006).

UCB-V
Text...

See: Ref (2006).

PAC-UCB
Text...

See: Ref (2006).

Gaussian UCB
Text...

See: Ref (2006).

Exp3
Text...

See: Ref (2006).

Meta-Bandits
Text...

See: Ref (2006).

Hierarchical Bandits
Text...

See: Ref (2006).



Node Initialisation

sdfg

History Heuristic
Text...

See: Ref (2006).

AMAF
Text...

See: Ref (2006).

RAVE
Text...

See: Ref (2006).

Transpositions
Text...

See: Ref (2006).

Reinforcement Learning
Text...

See: Ref (2006).

TDĪ±
Text...

See: Ref (2006).



Node Selection

sdfg

First Play Urgency
Text...

See: Ref (2006).

Move Pruning
Text...

See: Ref (2006).

Contextual MCTS
Text...

See: Ref (2006).

Nested MCTS
Text...

See: Ref (2006).

Regret Minimisation
Text...

See: Ref (2006).

Exploration Ordering
Text...

See: Ref (2006).

Move Groups
Text...

See: Ref (2006).

Multiobjective Methods
Text...

See: Ref (2006).

Sample Based Planning
Text...

See: Ref (2006).

Criticality
Text...

See: Ref (2006).

FAST
Text...

See: Ref (2006).

MAST
Text...

See: Ref (2006).

PAST
Text...

See: Ref (2006).

Exp3
Text...

See: Ref (2006).



Domain Specific

Playing strength can be improved by exploiting knowledge specific to the target domain (i.e game). This can lead to dramatic improvements in performance at the expense of generality.

Playout Heuristics
Text...

See: Ref (2006).

Domain Characteristics
Text...

See: Ref (2006).


Domain Independent

sdfg

Parameter Tuning
Text...

See: Ref (2006).

Forced Moves
Text...

See: Ref (2006).

Single Player MCTS (SP-MCTS)
Text...

See: Ref (2006).

Multiplayer MCTS
Text...

See: Ref (2006).

Realtime MCTS
Text...

See: Ref (2006).

Handling Uncertainty
Text...

See: Ref (2006).

TAG
Text...

See: Ref (2006).

RRTs
Text...

See: Ref (2006).

MHSP
Text...

See: Ref (2006).

UCT+
Text...

See: Ref (2006).

Cross-Entropy MCTS
Text...

See: Ref (2006).

Optimal Active Learning
Text...

See: Ref (2006).

Evolutionary Methods
Text...

See: Ref (2006).



Parallelisation

asfg

Root Parallelisation
Text...

See: Ref (2006).

Leaf Parallelisation
Text...

See: Ref (2006).

GPU Implementation
Text...

See: Ref (2006).