LostXine / naive-rl

Naive implementations of CSE525 HWs

naive-rl

Naive implementations of CSE525 HWs

Environment

Install

Please check Environment.ipynb

Test env

• Ubuntu 18.04
• Python 3.6.9
• torch 1.5.0
• numpy 1.18.3

HW1 [Done]

Algorithm

• Baseline: Vanilla Online Q Learning (without target network and replay buffer)
• Fitted Q Iteration
• DQN
• DQN-like-version of "SARSA"

Test env

• InvertedPendulumMuJoCoEnv-v0
• HalfCheetahMuJoCoEnv-v0
• Breakout-v4-ram

HW2 (working)

Algorithm

• REINFORCE
• Advantage Actor Critic (A2C)

Test env

• InvertedPendulumMuJoCoEnv-v0
• HalfCheetahMuJoCoEnv-v0

HW3 (working)

Algorithm

• Offline dynamics model learning
• Online dynamics model learning

Test env

• Breakout-v4