Skip to content

Monte Carlo Tree Search applied to Combinatorial Search Problems

License

Notifications You must be signed in to change notification settings

robfitzgerald/mcts

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

50 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MCTS

a Monte Carlo Tree Search implementation for combinatorial search, using a variation of the UCT cost function suitable for arbitrary cost functions.

UCT variant detailed in J.P. Pedroso and R. Rei, "Tree Search and Simulation", Applied Simulation and Optimization, Springer International Publishing, pp 119-131, 2015.

two versions concurrently offered in the repo:

  • cse.bdlab.fitzgero.mcts.MCTS2
    • most recently implemented
  • cse.bdlab.fitzgero.mcts.MonteCarloTreeSearch
    • older, works, inefficient

goals of MCTS2:

  • MCTS is an online anytime search and so for performance, it has been implemented using mutable semantics
  • operations in MCTS2 have been implemented as extension methods
  • state representation is a BitSet (see bitboard)
  • memory-specialized class representation for Leaf + Branch1 nodes and cost observations

License: MIT