search

Yoshi-0921 / MAEXP

Minimal frameworks for multi-agent reinforcement learning with deep neaural network.
MIT License
1 stars 0 forks source link

Noise robustness experiment 2 #14

Open Yoshi-0921 opened 3 years ago

Yoshi-0921 commented 3 years ago

動的ノイズ(学習)

確率的に２値が反転する環境

visible_range = 7

P={0.2, 0.1, 0.05} State
MAT-DQN done
Baseline done
P={0.5, 0.3, 0.1} State
MAT-DQN done
Baseline done
P={0.2, 0.0, 0.0} State
MAT-DQN done
Baseline done
P={0.5, 0.0, 0.0} State
MAT-DQN done
Baseline done
Yoshi-0921 commented 3 years ago

動的ノイズ(評価)

学習環境と評価環境のノイズ発生確率が異なると仮定。

visible_range = 7, P={0.0, 0.0, 0.0}

P={0.0, 0.0, 0.0} State
MAT-DQN done
Baseline done
P={0.2, 0.1, 0.05} State
MAT-DQN done
Baseline done
P={0.5, 0.3, 0.1} State
MAT-DQN done
Baseline done
P={0.2, 0.0, 0.0} State
MAT-DQN done
Baseline done
P={0.5, 0.0, 0.0} State
MAT-DQN done
Baseline done
Yoshi-0921 commented 3 years ago

動的ノイズ(評価)

学習環境と評価環境は同じと仮定。ノイズは同様に生成されると仮定。

visible_range = 7, P={0.2, 0.1, 0.05}

P={0.2, 0.1, 0.05} State
MAT-DQN
Baseline
P={0.5, 0.3, 0.1} State
MAT-DQN
Baseline
P={0.2, 0.0, 0.0} State
MAT-DQN
Baseline
P={0.5, 0.0, 0.0} State
MAT-DQN
Baseline
Yoshi-0921 commented 3 years ago

動的ノイズ(評価)

学習環境と評価環境は同じと仮定。ノイズは同様に生成されると仮定。

visible_range = 7, P={0.5, 0.3, 0.1}

P={0.2, 0.1, 0.05} State
MAT-DQN
Baseline
P={0.5, 0.3, 0.1} State
MAT-DQN
Baseline
P={0.2, 0.0, 0.0} State
MAT-DQN
Baseline
P={0.5, 0.0, 0.0} State
MAT-DQN
Baseline
Yoshi-0921 commented 3 years ago

動的ノイズ(評価)

学習環境と評価環境は同じと仮定。ノイズは同様に生成されると仮定。

visible_range = 7, P={0.2, 0.0, 0.0}

P={0.2, 0.1, 0.05} State
MAT-DQN
Baseline
P={0.5, 0.3, 0.1} State
MAT-DQN
Baseline
P={0.2, 0.0, 0.0} State
MAT-DQN
Baseline
P={0.5, 0.0, 0.0} State
MAT-DQN
Baseline
Yoshi-0921 commented 3 years ago

動的ノイズ(評価)

学習環境と評価環境は同じと仮定。ノイズは同様に生成されると仮定。

visible_range = 7, P={0.5, 0.0, 0.0}

P={0.0, 0.0, 0.0} State
MAT-DQN done
Baseline done
P={0.2, 0.1, 0.05} State
MAT-DQN done
Baseline done
P={0.5, 0.3, 0.1} State
MAT-DQN done
Baseline done
P={0.2, 0.0, 0.0} State
MAT-DQN done
Baseline done
P={0.5, 0.0, 0.0} State
MAT-DQN done
Baseline done