RL-A3C-Riverraid-v0

A simple implementation of A3C for atari game Riverraid-v0 from scratch. Joint work with @xmty777

• Only use torch data structure and some commonly used torch functions.
• Implement my own Conv2d, Linear and LSTM layers, including forward and backward.

How to train

  virtualenv --python=python3 venv
  source venv/bin/activate
  
  # install dependencies
  pip install -r requirement.txt
  
  # train
  python A3C/my_main.py

The detail hyperparameters are configed by python class Config in my_main.py. Train model and log would be stored in ./model and ./log separately.

How to test

No implementation. By you can implement your test file refer to PB17111656.py.

Result

Reference

pytorch-a3c

torch.optim.adam

Welcome to Deep Reinforcement Learning Part 1 : DQN

百度百科：卷积神经网络

pytorch官方文档：nn.functional

机器学习课件/Lec10.pdf

空洞卷积理解

怎样通俗易懂地解释反卷积？

Policy Gradient Methods for Reinforcement Learning with Function Approximation

AC、A2C、A3C算法

Actor-Critic

深度强化学习算法 A3C