A Stochastic Quasi-Newton Method for Large-Scale Nonconvex Optimization with Applications

TL;DR

This paper introduces a stochastic quasi-Newton method based on a damped and regularized BFGS approach, designed to efficiently solve large-scale nonconvex optimization problems.

Contribution

It presents a new stochastic quasi-Newton algorithm with damping and regularization for large-scale nonconvex optimization, extending classical BFGS methods.

Findings

Demonstrates convergence properties of the proposed method

Shows improved performance over existing stochastic optimization algorithms

Applicable to large-scale nonconvex problems in practice

Abstract

This paper proposes a novel stochastic version of damped and regularized BFGS method for addressing the above problems.