Comparison DQN variations with pytorch. MountainCar-v0, LunarLander-v2 is used for experiment.
- pytorch
- gym
- tensorboard
- numpy
Rewards per episodes training.
DQN: Orange
Double DQN: Blue
Rewards per episodes training.
DQN: Orange
Double DQN: Blue
Dueling DQN: Pink
D3QN: Silver
- DQN
- Double DQN
- Dueling DQN
- Dueling Double DQN
- PER
- NoiseNet
- C51
- Rainbow
- CNN