This is a library that the user can implement simple concurrent MCTS with small locking. This archived version is used on KAPO Science War(KAIST-POSTECH 학생대제전). After, it is being developed on the personal repository.
- Run MCTS rollouts concurrently(now use locking on creating node, will be improved to lock free).
- Use UCT policy with dynamic coefficient by
choose_weight
onAction
. - You can set the rewards on game over(
game_over
) and end of max searching depth(partial_rewards
)(it can apply onCyclePolicy::PartialReward
) onState
. - You can select cycle policy that allows cycle, ignores cycle and just do it, or use partial rewards.