Random Network Distillation

Intrinsic Reward Graph with play

Venture	Montezuma's Revenge
~	New model for Montezuma

• [x] Advantage Actor critic [1]
• [x] Parallel Advantage Actor critic [2]
• [x] Exploration by Random Network Distillation [3]
• [x] Proximal Policy Optimization Algorithms [4]

1. Setup

Requirements

• python3.6
• gym
• OpenCV Python
• PyTorch
• tensorboardX

2. How to Train

Modify the parameters in config.conf as you like.

python train.py

3. How to Eval

python eval.py

4. Loss/Reward Graph

• Montezuma's Revenge Env
• Venture Env

References

[1] Actor-Critic Algorithms
[2] Efficient Parallel Methods for Deep Reinforcement Learning
[3] Exploration by Random Network Distillation
[4] Proximal Policy Optimization Algorithms