Convergence Analysis of Gradient Descent for Eigenvector Computation

Zhiqiang Xu, Xin Cao, Xin Gao

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Scopus citations

Abstract

We present a novel, simple and systematic convergence analysis of gradient descent for eigenvector computation. As a popular, practical, and provable approach to numerous machine learning problems, gradient descent has found successful applications to eigenvector computation as well. However, surprisingly, it lacks a thorough theoretical analysis for the underlying geodesically non-convex problem. In this work, the convergence of the gradient descent solver for the leading eigenvector computation is shown to be at a global rate O(min{ (lambda_1/Delta_p)^2 log(1/epsilon), 1/epsilon }), where Delta_p=lambda_p-lambda_p+1>0 represents the generalized positive eigengap and always exists without loss of generality with lambda_i being the i-th largest eigenvalue of the given real symmetric matrix and p being the multiplicity of lambda_1. The rate is linear at (lambda_1/Delta_p)^2 log(1/epsilon) if (lambda_1/Delta_p)^2=O(1), otherwise sub-linear at O(1/epsilon). We also show that the convergence only logarithmically instead of quadratically depends on the initial iterate. Particularly, this is the first time the linear convergence for the case that the conventionally considered eigengap Delta_1= lambda_1 - lambda_2=0 but the generalized eigengap Delta_p satisfies (lambda_1/Delta_p)^2=O(1), as well as the logarithmic dependence on the initial iterate are established for the gradient descent solver. We are also the first to leverage for analysis the log principal angle between the iterate and the space of globally optimal solutions. Theoretical properties are verified in experiments.
Original languageEnglish (US)
Title of host publicationProceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence
PublisherInternational Joint Conferences on Artificial Intelligence
ISBN (Print)9780999241127
DOIs
StatePublished - Jul 5 2018

Fingerprint Dive into the research topics of 'Convergence Analysis of Gradient Descent for Eigenvector Computation'. Together they form a unique fingerprint.

Cite this