NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:4667
Title:Who is Afraid of Big Bad Minima? Analysis of gradient-flow in spiked matrix-tensor models

This paper analyzes the critical points in the spiked matrix-tensor model, framing it as a proxy for more general models in high-dimensional inference. It examines the signal to noise ratio, its effect on spurious local minima, and the behavior of gradient flow. I agree with the reviewers' consensus that this non-trivial analysis would be interesting to the NeurIPS community and I recommend acceptance.