Noisy Gradient Descent Converges to Flat Minima for Nonconvex Matrix Factorization.

Tianyi Liu, Yan Li, Song Wei, Enlu Zhou, Tuo Zhao

Browse the full AISTATS paper archive.