Exit Time Analysis for Approximations of Gradient Descent Trajectories Around Saddle Points

Abstract

This paper considers the problem of understanding the exit time for trajectories of gradient-related first-order methods from saddle neighborhoods under some initial boundary conditions. Given the 'flat' geometry around saddle points, first-order methods can struggle to escape these regions in a fast manner due to the small magnitudes of gradients encountered. In particular, while it is known that gradient-related first-order methods escape strict-saddle neighborhoods, existing analytic techniques do not explicitly leverage the local geometry around saddle points in order to control behavior of gradient trajectories. It is in this context that this paper puts forth a rigorous geometric analysis of the gradient-descent method around strict-saddle neighborhoods using matrix perturbation theory. In doing so, it provides a key result that can be used to generate an approximate gradient trajectory for any given initial conditions. In addition, the analysis leads to a linear exit-time solution for gradient-descent method under certain necessary initial conditions, which explicitly bring out the dependence on problem dimension, conditioning of the saddle neighborhood, and more, for a class of strict-saddle functions.

arXiv comment70 pages; pre-print of the journal paper published in Information and Inference: A Journal of the IMA, 2023

References in corpus (13)

• Phase Retrieval via Wirtinger Flow: Theory and Algorithms
• Implicit Regularization in Nonconvex Statistical Estimation: Gradient Descent Converges Linearly for Phase Retrieval, Matrix Completion, and Blind Deconvolution
• Calculus of the exponent of Kurdyka-Łojasiewicz inequality and its applications to linear convergence of first-order methods
• Non-convex learning via Stochastic Gradient Langevin Dynamics: a nonasymptotic analysis
• STFT Phase Retrieval: Uniqueness Guarantees and Recovery Algorithms
• Gradient Descent with Random Initialization: Fast Global Convergence for Nonconvex Phase Retrieval
• Gradient Descent Converges to Minimizers
• Escaping Saddles with Stochastic Gradients
• First-order Methods Almost Always Avoid Saddle Points
• Accelerated Gradient Descent Escapes Saddle Points Faster than Gradient Descent
• A Generic Approach for Escaping Saddle points
• Eigenvector Under Random Perturbation: A Nonasymptotic Rayleigh-Schrödinger Theory
• On the fast convergence of random perturbations of the gradient flow

Cited by in corpus (2)

• A Linearly Convergent Algorithm for Distributed Principal Component Analysis
• Accelerated Gradient Methods for Nonconvex Optimization: Escape Trajectories From Strict Saddle Points and Convergence to Local Minima