Local LMO: Constrained Gradient Optimization via a Local Linear Minimization Oracle
Peter RichtárikKAUSTKaja Gruntkowska · 2026
Read Local LMO: Constrained Gradient Optimization via a Local Linear Minimization Oracle by Peter RichtárikKAUSTKaja Gruntkowska (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-08850.txt.
Introduction
Peter Richtárik, Kaja Gruntkowska, and Hanmin Li set out Broximal Gradient Descent, a bound task plan that needs no cast back step. The cast back step of its kin plan can cost a lot, or sit ill with hard bound sets. BroxGD swaps it for a brox step. Each step finds the best of a flat guess score. The hunt stays in the bound set, cut with a ball round the live point. The ball's size is set by rule.
The team builds a wide proof of its pace, cross many task kinds and size rules. Toy cases show how it moves. Bound work gains a plan whose steps stay in bounds by build.
Catalog record
| CREATOR | Peter RichtárikKAUSTKaja Gruntkowska |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 189 KB |
| LICENSE | http://creativecommons.org/publicdomain/zero/1.0/ |
| RIGHTS | CC0 |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.08850