Nesterov’s Accelerated Gradient for Unbounded Convex Functions Finds the Minimum-Norm Point in the Dual Space

Keiya Sakabe · 2026

Read Nesterov’s Accelerated Gradient for Unbounded Convex Functions Finds the Minimum-Norm Point in the Dual Space by Keiya Sakabe (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-2602-08618.txt.

Description

Gradient descent on a function with no lower bound diverges, and that divergence itself carries information. Keiya Sakabe studies first-order methods on lower-unbounded convex functions, where trajectories of gradient descent and Nesterov's accelerated method both run away. His arXiv paper quantifies their speeds and directions of escape, with implications for judging unboundedness.

The key is a dual norm-minimization problem, solvable by mirror descent with the conjugate function as its distance generator. Gradient descent on the original problem coincides with mirror descent on the dual, solving both at a one-over-k rate. Nesterov's method, unmodified, accelerates both to a one-over-k-squared rate. Readers get a precise account of divergence as dual optimization in disguise.

Catalog record

CREATORKeiya Sakabe
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE108 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-08618

Browse the full library · About the Tower of Babel