Dynamical Low-Rank Compression of Neural Networks with Robustness under Adversarial Attacks

Steffen Schotthöfer, H. Lexie Yang, and Stefan Schnake · 2025

Read Dynamical Low-Rank Compression of Neural Networks with Robustness under Adversarial Attacks by Steffen Schotthöfer, H. Lexie Yang, and Stefan Schnake (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-2505-08022.txt.

Description

Small devices need small neural networks, but squeezed models tend to fold under adversarial inputs, tiny crafted changes that fool a network. Steffen Schotthöfer and co-authors tackle both pressures at once in a 2025 paper on low-rank compression.

Their training scheme keeps each layer in a compact low-rank form while a new spectral penalty controls a stability number for the compressed core. Rank adapts on its own, so each network compresses to the size it can bear. Across standard models, data sets, and attack types, the regularized networks shrank by more than ninety-four percent. They also matched or beat the uncompressed originals on adversarial accuracy, without losing ground on clean data.

Catalog record

CREATORSteffen Schotthöfer, H. Lexie Yang, and Stefan Schnake
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE68 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-08022

Browse the full library · About the Tower of Babel