tianshou
Tianshou(天授) is a reinforcement learning platform.
data
TODO:
Replay Memory
Multiple wirter/reader
Importance sampling
simulator
go(for AlphaGo)
environment
gym
core
TODO:
Optimizer
MCTS
agent (optional)
DQNAgent etc.
Pontential Bugs:
- Wrong calculation of eval value
UCTNode.cpp
106 if (to_move == FastBoard::WHITE) {
107 net_eval = 1.0f - net_eval;
108 }
309 if (tomove == FastBoard::WHITE) {
310 score = 1.0f - score;
311 }
- create children only on leaf node
UCTSearch.cpp
60 if (!node->has_children() && m_nodes < MAX_TREE_SIZE) {
61 float eval;
62 auto success = node->create_children(m_nodes, currstate, eval);
63 if (success) {
64 result = SearchResult(eval);
65 }
66 }
Description
Languages
Python
100%