NeurIPS 2020

Entrywise convergence of iterative methods for eigenproblems


Meta Review

Overall the reviewers found that an improvement to such a fundamental, well-used algorithm, like the power method, was surprising and interesting, and the analysis for l2-linf norm was interesting. There were some weaknesses concerning the assumptions, but post-rebuttal the reviewers seemed to agree that Assumption 2 was not needed actually, and this caused a reviewer to think the proof was more complete / natural.