Landscape Complexity for the Empirical Risk of Generalized Linear Models: Discrimination between Structured Data

Theodoros G. Tsironis, Aris L. Moustakas · 2025

Read Landscape Complexity for the Empirical Risk of Generalized Linear Models: Discrimination between Structured Data by Theodoros G. Tsironis, Aris L. Moustakas (2025) 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-2503-14403.txt.

Description

Theodoros Tsironis and Aris Moustakas count the critical points of loss surfaces for simple learning models, in a 2025 paper. The count gauges how rugged the training landscape is. Data in real systems often carry structure, shown here as links between input parts. The team works in the limit where data size and input size grow together at a fixed ratio.

Formulas from random matrix theory give the average count at each loss level. The single teaching unit case comes first, in full detail. Then two rival data sets with unlike link patterns enter, and the unit must tell them apart. That case speaks to how hostile or shaped data bend the landscape. Results are exact in the large-size limit, under one stated premise.

Catalog record

CREATORTheodoros G. Tsironis, Aris L. Moustakas
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE54 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-14403

Browse the full library · About the Tower of Babel