I want to ask you if it's possible by using neural networks jointly with the Contextual Bandit algorithm to learn the probability distributions by which the rewards are computed as a function of the agent's actions and the state in which the environment is ? I've seen many examples of Contextual Bandit, some using neural networks as function approximators, but the probability distributions were always provided and usually Thompson sampling is used as a mean to balance exploitation and exploration. But I wonder if it's possible somehow to use two neural networks in parallel, one for learning the probability distributions and the other as a function approximator.

I'm new to Reinforcement Learning so please forgive me if my question is basic and doesn't suit this forum, I'll delete it as soon as it's labeled as such.


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.