English

Hybrid Deterministic-Stochastic Methods for Data Fitting

Numerical Analysis 2018-08-23 v4 Systems and Control Optimization and Control Machine Learning

Abstract

Many structured data-fitting applications require the solution of an optimization problem involving a sum over a potentially large number of measurements. Incremental gradient algorithms offer inexpensive iterations by sampling a subset of the terms in the sum. These methods can make great progress initially, but often slow as they approach a solution. In contrast, full-gradient methods achieve steady convergence at the expense of evaluating the full objective and gradient on each iteration. We explore hybrid methods that exhibit the benefits of both approaches. Rate-of-convergence analysis shows that by controlling the sample size in an incremental gradient algorithm, it is possible to maintain the steady convergence rates of full-gradient methods. We detail a practical quasi-Newton implementation based on this approach. Numerical experiments illustrate its potential benefits.

Keywords

Cite

@article{arxiv.1104.2373,
  title  = {Hybrid Deterministic-Stochastic Methods for Data Fitting},
  author = {Michael P. Friedlander and Mark Schmidt},
  journal= {arXiv preprint arXiv:1104.2373},
  year   = {2018}
}

Comments

26 pages. Revised proofs of Theorems 2.6 and 3.1, results unchanged

R2 v1 2026-06-21T17:53:16.610Z