The paper considers the problem of operator valued indefinite kernels. It presents theoretical analysis, optimization method, and empirical results. Four knowledgeable reviewers carefully considered the paper, and all of them recommended accept. Therefore I also recommend to accept the paper at NeurIPS 2020. The authors are encouraged to address all the issues raised by the reviewers for the final version. For example the comparison of stabilization and minimization by Reviewer #1 and #4, and overall to improve the motivation of using operator valued indefinite kernels.