Algorithmic Regularization in Over-parameterized Matrix Sensing and Neural Networks with Quadratic Activations

12/26/2017 ∙ by Yuanzhi Li, et al. ∙ 0

We show that the (stochastic) gradient descent algorithm provides an implicit regularization effect in the learning of over-parameterized matrix factorization models and one-hidden-layer neural networks with quadratic activations. Concretely, we show that given Õ(dr^2) random linear measurements of a rank r positive semidefinite matrix X^, we can recover X^ by parameterizing it by UU^ with U∈R^d× d and minimizing the squared loss, even if r ≪ d. We prove that starting from a small initialization, gradient descent recovers X^ in Õ(√(r)) iterations approximately. The results solve the conjecture of Gunasekar et al.'17 under the restricted isometry property. The technique can be applied to analyzing neural networks with quadratic activations with some technical modifications.



There are no comments yet.


page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.