I am running a Deep RL algorithm. I defined a custom reward function. I run the algorithm for at least 500 epochs.

For each epoch, I am printing the total reward received by the actor-network. It is around $- 10^5$ for the first epoch. After the completion of 500 epochs, the value is almost similar.

I ran with the hope that after enough epochs, the value decreases and move toward zero. But my guess went wrong.

What can I infer from his? Is my actor learning? If yes, then why do the total reward values not decrease?

Note that I want to infer the learning of my actor network from total reward per epoch only and not from its actions in the environment.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy