Id: arxiv/2602.20646 Url: https://arxiv.org/abs/2602.20646 Tags: math.OC, cs.LG, stat.ML Score: 0 Title: On the Convergence of Stochastic Gradient Descent with Perturbed Forward-Backward Passes Domain: math.OC Source: arxiv Authors: Boao Kong, Hengrui Zhang, Kun Yuan Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-02-24 Page: https://delphi.science/r/arxiv/2602.20646 We study stochastic gradient descent (SGD) for composite optimization problems with $N$ sequential operators subject to perturbations in both the forward and backward passes. Unlike classical analyses that treat gradient noise as additive and localized, perturbations to intermediate outputs and gradients cascade through the computational graph, compounding geometrically with the number of operators. We present the first comprehensive theoretical analysis of this setting. Specifically, we characterize how forward and backward perturbations propagate and amplify within a single gradient step,…