Skip to content

Add best response using Monte Carlo Tree Search

Jan Maděra requested to merge maderja1/GTLib2:MCTSBestResponse into master

Added MonteCarloTreeSearch to CMakeLists.txt source files.

implemented Monte Carlo Tree Search is a basic version of MCTS without any enhancements.

Instead of BehavioralStrategy new type Strategy was used which contains only the action with the highest utility or the most explored action.

MCTS is divided into two parts which are default policy and tree policy. Default policy selects an expandable node and expands it. Tree policy simulates a random game and backpropagates the simulation result.

Merge request reports