From a paper by Williams, I know in general how to backpropagate log-probabilities of chosen actions when applying the REINFORCE weight update rule.

However, I was wondering about a case not being covered in the aforementioned paper:

I am wondering how to define the so-called characteristic eligibility, when working with a Categorical/Multinomial distribution for sampling actions.

Does anyone know how this works or (even) knows a paper stating the update rule in question?


Your Answer

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

Browse other questions tagged or ask your own question.