Delphi.science

Proximal random reshuffling with inexact gradients for nonconvex composite optimization

Source: arxiv.org/abs/2611.15916

Boris S. Mordukhovich, Pedro Pérez-Aros

arXivarxiv/2611.15916Nov 11, 2026math.OCscore 0

We design and justify a proximal inexact random reshuffling method for minimizing sums of the finite average of functions with Lipschitz continuous gradients and lower semicontinuous regularizers that may be nonconvex. Each epoch visits every component once, using inexact gradients with a prescribed tolerance, and ends with a single proximal step. Our analysis is pathwise: it holds for every sequence of permutations and does not depend on how the reshuffling is performed. We show that when the stepsizes and error tolerances satisfy suitable summability conditions, the objective values…

JSONText

Comments and reviews

No comments yet.

Supports Markdown, links, code fences, inline LaTeX ($x$ or \(x\)), and display LaTeX ($$...$$ or \[...\]).

Drafts stay in this tab.

API

curl /api/v1/records/arxiv/2611.15916
curl -X POST /api/v1/records/arxiv/2611.15916/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2611.15916/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'