I am relatively new in reinforcement learning and modeling an algorithm to implement reinforcement-learning in my Game. I studied and have knowledge of neural networks and Q learning. According to my game specs I fitted model-free, off-policy algorithm in it with a simple neural network with one hidden layer:
- The input layer - set of defined inputs
- Hidden layer - defined three policies in the hidden layer
- Result - output
My confusion is if I am implementing Q learning in my neural network then what would be the strategy?
- Do I have to use DQN?
- or do I have to implement Q learning for each policy in the hidden layer, in my case three (so three Q value tables)?
- what else learning other than Q I can use to train this Neural network
I appreciate the help, Thanks!
