Acceleration for Polyak-Łojasiewicz Functions with a Gradient Aiming Condition
Julien Hermant · 2026
Read Acceleration for Polyak-Łojasiewicz Functions with a Gradient Aiming Condition by Julien Hermant (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-2602-10022.txt.
Description
Momentum methods famously speed up gradient descent for strongly convex problems, but their edge fades for a wider class of functions. A 2026 paper by Julien Hermant examines minimization of smooth Polyak-Łojasiewicz functions, where momentum was thought to help via an intermediate class called strongly quasar-convex functions. Hermant shows that extra structure does not suffice in general.
The paper proves that strong quasar-convexity alone cannot guarantee better worst-case bounds for momentum over plain gradient descent. As an alternative, Hermant studies these functions under an aiming condition, which gauges whether the descent direction aims at a minimizer. That geometric view clarifies what actually enables provable acceleration by momentum.
Catalog record
| CREATOR | Julien Hermant |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 123 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-10022