It seems the contextual bandits problems, as shown in these two papers Tong Mu, Yash Chandak, Tatsunori Hashimoto, Emma Brunskill: Factored DRO: Factored Distributionally Robust Policies for Contextual Bandits, and Nian Si, Fan Zhang, Zhengyuan Zhou, Jose Blanchet: Distributionally Robust Batch Contextual Bandits, generally assumes the reward $R$ are bounded $0<R<M$. I am wondering what would be impacted if $R$ is allowed to be negative.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.