I am following the OpenAI's spinning up tutorial Part 3: Intro to Policy Optimization. It is mentioned there that the reward-to-go reduces the variance of the policy gradient. While I understand the intuition behind it, I struggle to find a proof in the literature.
$\begingroup$
$\endgroup$
-
1$\begingroup$ Does the answer to this question answer yours as well? $\endgroup$ – user5093249 Jun 10 '20 at 13:55
-
$\begingroup$ No, the linked question only proofs that the reward-to-go does not introduce any bias to the gradient estimate. $\endgroup$ – sirKris van Dela Jun 10 '20 at 14:14
-
$\begingroup$ This is nontrivial to prove, actually anything involving stochastic function approximation is nontrivial. You can search research papers, you won't find it in any book right now $\endgroup$ – FourierFlux Jun 10 '20 at 14:33
