9713faa264b94e2bf346a1bb52587fd8-AuthorFeedback.pdf

We thank the reviewers for their valuable comments and for pointing out the typos and clarity issues. We first explain the novelty and significance of our work and then answer the questions of each reviewer. We will clarify our contributions in the revision. But more careful analysis are needed to obtain similar convergence rate without projection. Here we adopt Gaussian policy for simplicity.