Guardado en:
Detalles Bibliográficos
Autores principales: Tadipatri, Uday Kiran Reddy, Haeffele, Benjamin D., Agterberg, Joshua, Vidal, René
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2411.02767
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912281875447808
author Tadipatri, Uday Kiran Reddy
Haeffele, Benjamin D.
Agterberg, Joshua
Vidal, René
author_facet Tadipatri, Uday Kiran Reddy
Haeffele, Benjamin D.
Agterberg, Joshua
Vidal, René
contents 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 and sensing, single-layer multi-head attention mechanisms, tensor factorization, deep linear and ReLU networks, and more. Our general framework is based on linking the non-convex empirical risk minimization (ERM) problem to a closely related convex optimization problem over prediction functions, which provides a global, achievable lower-bound to the 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 almost linearly with the network width.
format Preprint
id arxiv_https___arxiv_org_abs_2411_02767
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle A Convex Relaxation Approach to Generalization Analysis for Parallel Positively Homogeneous Networks
Tadipatri, Uday Kiran Reddy
Haeffele, Benjamin D.
Agterberg, Joshua
Vidal, René
Machine Learning
Signal Processing
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 and sensing, single-layer multi-head attention mechanisms, tensor factorization, deep linear and ReLU networks, and more. Our general framework is based on linking the non-convex empirical risk minimization (ERM) problem to a closely related convex optimization problem over prediction functions, which provides a global, achievable lower-bound to the 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 almost linearly with the network width.
title A Convex Relaxation Approach to Generalization Analysis for Parallel Positively Homogeneous Networks
topic Machine Learning
Signal Processing
url https://arxiv.org/abs/2411.02767