# A path-norm toolkit for modern networks: consequences, promises and challenges

Source: [A path-norm toolkit for modern networks: consequences, promises and challenges](<https://www.sota2.com/research/paper/68fa5cae1e209131ce9d3499>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-25T17:47:19.731Z

Authors: Antoine Gonon, Nicolas Brisebarre, Elisa Riccietti, R\\'emi Gribonval

Published: 2023-10-02T14:12:53.000Z

## Abstract

This work introduces the first toolkit around path-norms that fully encompasses general DAG ReLU networks with biases, skip connections and any operation based on the extraction of order statistics: max pooling, GroupSort etc. This toolkit notably allows us to establish generalization bounds for modern neural networks that are not only the most widely applicable path-norm based ones, but also recover or beat the sharpest known bounds of this type. These extended path-norms further enjoy the usual benefits of path-norms: ease of computation, invariance under the symmetries of the network, and improved sharpness on layered fully-connected networks compared to the product of operator norms, another complexity measure most commonly used.   The versatility of the toolkit and its ease of implementation allow us to challenge the concrete promises of path-norm-based generalization bounds, by numerically evaluating the sharpest known bounds for ResNets on ImageNet.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2310.01225>)

[PDF](<https://arxiv.org/pdf/2310.01225.pdf>)