Building a deep reinforcement learning with a cnn q - approximation

Question

I'm new in DRL. Starting from this code https://github.com/jaromiru/cwcf, I would like to substitute the MLP used for the q function approximation with a CNN, but I don't know how to do. Can anybody help me? Thanks

Answer 1

Try going through this it has a detailed explanation on how to build DQN to solve the CartPole problem. You can also have a look at this which has implementations of many DRL algorithms

Then you can replace the code in agent.py present in repo with DQN agent code