Provable Efficient Online Matrix Completion via Non-convex Stochastic Gradient Descent

TL;DR

This paper introduces the first provable, efficient online matrix completion algorithm using non-convex stochastic gradient descent, enabling real-time updates with near linear runtime and competitive accuracy.

Contribution

It develops a novel online matrix completion method with theoretical guarantees, extending non-convex SGD techniques to the online setting for the first time.

Findings

Algorithm achieves near linear runtime per update.

Provides competitive sample complexity compared to offline methods.

Introduces a framework for analyzing SGD avoiding saddle points in non-convex problems.

Abstract

Matrix completion, where we wish to recover a low rank matrix by observing a few entries from it, is a widely studied problem in both theory and practice with wide applications. Most of the provable algorithms so far on this problem have been restricted to the offline setting where they provide an estimate of the unknown matrix using all observations simultaneously. However, in many applications, the online version, where we observe one entry at a time and dynamically update our estimate, is more appealing. While existing algorithms are efficient for the offline setting, they could be highly inefficient for the online setting. In this paper, we propose the first provable, efficient online algorithm for matrix completion. Our algorithm starts from an initial estimate of the matrix and then performs non-convex stochastic gradient descent (SGD). After every observation, it performs a fast update involving only one row of two tall matrices, giving near linear total runtime. Our algorithm can be naturally used in the offline setting as well, where it gives competitive sample complexity and runtime to state of the art algorithms. Our proofs introduce a general framework to show that SGD updates tend to stay away from saddle surfaces and could be of broader interests for other non-convex problems to prove tight rates.