Duality Based Generalization Bounds for Parallel Positively Homogeneous Networks

Uday Kiran Reddy Tadipatri, Benjamin David Haeffele, Joshua Agterberg, Rene Vidal

Keywords: statistical learning theory, non-convex optimization, high dimension, sample complexity, matrix sensing, neural network, positively homogeneous, generalization

TL;DR: A general framework for deriving generalization bounds for parallel positively homogeneous neural networks.

Abstract:

We propose a general framework for deriving generalization bounds for parallel positively homogeneous neural
networks--a class of neural networks whose input-output map decomposes as the sum of positively homogeneous maps.
Examples of such networks include matrix factorization/sensing, two-layer linear/ReLU networks,
single-layer multi-head attention mechanisms, and tensor factorization.
Our general framework is based on linking the non-convex empirical risk minimization (ERM) problem of interest to a
closely related convex optimization problem over prediction functions, which provides a global, achievable
lower-bound to the original ERM problem. We exploit this convex lower-bound to perform generalization
analysis in the convex space while controlling the discrepancy between the convex model and its non-convex
counterpart. We apply our general framework to a wide variety of models ranging from low-rank matrix sensing,
to structured matrix sensing, two-layer linear networks, two-layer ReLU networks, and single-layer
multi-head attention mechanisms, achieving generalization bounds with a sample complexity that scales nearly
linear with the network width.

Primary Area: learning theory

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 12442