55 lines
961 B
Markdown
55 lines
961 B
Markdown
# 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:
|
|
0. 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 }
|
|
|
|
1. 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 }
|
|
|
|
|
|
|