End-to-end-DRL-for-FJSP
End-to-end-DRL-for-FJSP copied to clipboard
PPO Critic
@leikun-starting Thank you very much!
I have some questions in regard to the critic, maybe from there some improvements arise.
- Currently the critic only receives the job embedding and not the machine embedding
- Why are you not updating the critic separately, e.g. with the job_v_loss
job_v_loss = self.MSE(val[j], rewards_all_env[j])
currently you simply add the critic loss to the actor loss. - The critic is used in both loss functions (job and machine)