Ada Lovelace

The way this code is written you would not be able to access the model outside of the class object. There would be many ways we could about this but...

1. There is a function updating the target model. Does this answer your question? 2. How about, state = [[a, b, c]] ?

ddqn is double here since there is no dueling implementation in the repo. The hyperparameters affect the performance of a model greatly. When comparing implementations please make sure to fix...

1. Loading just the weights of the model then sampling from a new set of experiences may be leading to divergence. 2. How about saving after each episode? How about...

The action shouldn't be restricted since it's the goal of the agent to learn what action to take for the most reward.

You would use the action associated with the most reward that is returned when you feed the state into the model.

The behavior we want from the Neural Network is to give us a similar output when given a similar state. As for the epoch, the model is constantly retrained after...

Did this ever make it to a release?

@dwSun @danqing I would also like to know the result of this as well. My 1060 is looking to be twice as fast than my Xeon Phi on a different...