A Retrospective Approximation Approach for Smooth Stochastic Optimization

• 成果类型: Article
• 署名作者: Newton, David; Bollapragada, Raghu; Pasupathy, Raghu; Yip, Nung Kwan
• 署名单位: Purdue University System; Purdue University; University of Texas System; University of Texas Austin; Indian Institute of Technology System (IIT System); Indian Institute of Technology (IIT) - Madras; Purdue University System; Purdue University
• 刊物名称: MATHEMATICS OF OPERATIONS RESEARCH
• ISSN/ISSBN: 0364-765X
• DOI: 10.1287/moor.2022.0136
• 发表日期: 2025
• 关键词: INEQUALITIES; RISK; sums

摘要

Stochastic Gradient (SG) is the de facto iterative technique to solve stochastic optimization (SO) problems with a smooth (nonconvex) objective f and a stochastic first-order oracle. SG's attractiveness is due in part to its simplicity of executing a single step along the negative subsampled gradient direction to update the incumbent iterate. In this paper, we question SG's choice of executing a single step as opposed to multiple steps between subsample updates. Our investigation leads naturally to generalizing SG into Retrospective Approximation (RA), where, during each iteration, a deterministic solver executes possibly multiple steps on a subsampled deterministic problem and stops when further solving is deemed unnecessary from the standpoint of statistical efficiency. RA thus formalizes what is appealing for implementation-during each iteration, plug in a solver-for example, L-BFGS line search or Newton-CG-as is, and solve only to the extent necessary. We develop a complete theory using relative error of the observed gradients as the principal object, demonstrating that almost sure and L1 consistency of RA are preserved under especially weak conditions when sample sizes are increased at appropriate rates. We also characterize the iteration and oracle complexity (for linear and sublinear solvers) of RA and identify a practical termination criterion leading to optimal complexity rates. To subsume nonconvex f, we present a certain random central limit theorem that incorporates the effect of curvature across all first-order critical points, demonstrating that the asymptotic behavior is described by a certain mixture of normals. The message from our numerical experiments is that the ability of RA to incorporate existing second-order deterministic solvers in a strategic manner might be important from the standpoint of dispensing with hyper-parameter tuning.