MCTS

E748468

MCTS is a heuristic search algorithm that uses randomized simulations to efficiently explore large decision trees, widely applied in game-playing AI and other complex planning problems.

Try in SPARQL Jump to: Statements Referenced by

Statements (50)

Predicate Object
instanceOf Monte Carlo method
heuristic search algorithm
advantage can be interrupted at any time with best-so-far move
does not require evaluation function a priori
handles large branching factors
canBeCombinedWith domain knowledge
heuristic evaluation functions
neural networks
coreStep backpropagation
expansion
selection
simulation
designedFor large decision spaces
sequential decision-making problems
fullName Monte Carlo Tree Search NERFINISHED
goal approximate optimal decisions
balance exploration and exploitation
hasVariant Nested Monte Carlo Search NERFINISHED
Parallel MCTS NERFINISHED
Progressive Widening NERFINISHED
Rapid Action Value Estimation NERFINISHED
Upper Confidence bounds applied to Trees NERFINISHED
input game state
transition model or simulator
limitation high computational cost for deep search
performance depends on simulation quality
operatesOn decision trees
output action value estimates
recommended action
property anytime algorithm
asymmetric tree growth
does not require full tree expansion
model-free with respect to value function
relatedTo Markov decision processes NERFINISHED
bandit algorithms
reinforcement learning
selectionPolicy upper confidence bounds for trees
selectionPolicyAbbreviation UCT NERFINISHED
uses Monte Carlo rollouts
randomized simulations
statistical sampling
widelyUsedIn Go programs
chess engines
combinatorial optimization
game-playing artificial intelligence
general game playing
planning
real-time strategy games
robot motion planning
scheduling problems

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.