search

mpnunez / Connect4-AI

Training an AI Player to play Connect4
0 stars 0 forks source link

Implement double DQN #8

Closed mpnunez closed 2 months ago

mpnunez commented 3 months ago

Dueling DQN: Split Q value into the value of the state $V(state)$ and the advantage of each action $A(state,action)$

mpnunez commented 3 months ago

Double DQN: Use on-policy Q to select best action for next state, but use target network to compute its Q-value

mpnunez commented 2 months ago

Double DQN: https://github.com/mpnunez/Connect4-AI/commit/67c4dff0563f01d34988becacd3713a73ca1cea9 Deuling DQN: https://github.com/mpnunez/Connect4-AI/commit/423259eaffd4f8c1e867cdb34065659328da09d1