NeurIPS 2020

Learning Feature Sparse Principal Subspace

Meta Review

The reviewers all very much liked this paper. The few complaints regarding the condition number and Theorem 5.1 and the convergence analysis to a stationary point in Theorem 5.3, were adequately addressed in the rebuttal. The problem is fundamental, and it is surprising that the authors can demonstrate convergence and iteration complexity for this subclass of problems. The experiments are also very strong. Overall, solid accept.