Please enable JavaScript.

Coggle requires JavaScript to display documents.

DQN, Initialize and start (episode n start, if not the end (Initialize…

- - - - select action
        choose A
        
        perform action
        do A on S
        
        observe reward and new state
        get R+1 and S+1 #
        
        add the transition S, A, R+1, S+1
        into memory
        
        optimize our model;
        only if our memory is more than batch size.
        #
        
        is done? game over?
        S = S_t?
        
        2 more items...
- - - - optimize the model by doing a gradient descent