A Distributed Quasi-Newton Algorithm for Primal and Dual Regularized Empirical Risk Minimization

TL;DR

This paper introduces a distributed second-order optimization algorithm for empirical risk minimization that efficiently uses curvature information, improving convergence speed and reducing communication costs in both primal and dual settings.

Contribution

It presents a novel distributed quasi-Newton method that leverages global Hessian approximations, outperforming existing methods especially in dual ERM problems.

Findings

Significantly reduces communication costs compared to state-of-the-art methods.

Achieves global linear convergence for a wide range of ERM problems.

Demonstrates faster convergence and lower runtime in computational experiments.

Abstract

We propose a communication- and computation-efficient distributed optimization algorithm using second-order information for solving empirical risk minimization (ERM) problems with a nonsmooth regularization term. Our algorithm is applicable to both the primal and the dual ERM problem. Current second-order and quasi-Newton methods for this problem either do not work well in the distributed setting or work only for specific regularizers. Our algorithm uses successive quadratic approximations of the smooth part, and we describe how to maintain an approximation of the (generalized) Hessian and solve subproblems efficiently in a distributed manner. When applied to the distributed dual ERM problem, unlike state of the art that takes only the block-diagonal part of the Hessian, our approach is able to utilize global curvature information and is thus magnitudes faster. The proposed method enjoys global linear convergence for a broad range of non-strongly convex problems that includes the most commonly used ERMs, thus requiring lower communication complexity. It also converges on non-convex problems, so has the potential to be used on applications such as deep learning. Computational results demonstrate that our method significantly improves on communication cost and running time over the current state-of-the-art methods.