Saved in:
Bibliographic Details
Main Authors: Mishkin, Aaron, Sahiner, Arda, Pilanci, Mert
Format: Preprint
Published: 2022
Subjects:
Online Access:https://arxiv.org/abs/2202.01331
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912316733259776
author Mishkin, Aaron
Sahiner, Arda
Pilanci, Mert
author_facet Mishkin, Aaron
Sahiner, Arda
Pilanci, Mert
contents We develop fast algorithms and robust software for convex optimization of two-layer neural networks with ReLU activation functions. Our work leverages a convex reformulation of the standard weight-decay penalized training problem as a set of group-$\ell_1$-regularized data-local models, where locality is enforced by polyhedral cone constraints. In the special case of zero-regularization, we show that this problem is exactly equivalent to unconstrained optimization of a convex "gated ReLU" network with non-singular gates. For problems with non-zero regularization, we show that convex gated ReLU models obtain data-dependent approximation bounds for the ReLU training problem. To optimize the convex reformulations, we develop an accelerated proximal gradient method and a practical augmented Lagrangian solver. We show that these approaches are faster than standard training heuristics for the non-convex problem, such as SGD, and outperform commercial interior-point solvers. Experimentally, we verify our theoretical results, explore the group-$\ell_1$ regularization path, and scale convex optimization for neural networks to image classification on MNIST and CIFAR-10.
format Preprint
id arxiv_https___arxiv_org_abs_2202_01331
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Fast Convex Optimization for Two-Layer ReLU Networks: Equivalent Model Classes and Cone Decompositions
Mishkin, Aaron
Sahiner, Arda
Pilanci, Mert
Machine Learning
We develop fast algorithms and robust software for convex optimization of two-layer neural networks with ReLU activation functions. Our work leverages a convex reformulation of the standard weight-decay penalized training problem as a set of group-$\ell_1$-regularized data-local models, where locality is enforced by polyhedral cone constraints. In the special case of zero-regularization, we show that this problem is exactly equivalent to unconstrained optimization of a convex "gated ReLU" network with non-singular gates. For problems with non-zero regularization, we show that convex gated ReLU models obtain data-dependent approximation bounds for the ReLU training problem. To optimize the convex reformulations, we develop an accelerated proximal gradient method and a practical augmented Lagrangian solver. We show that these approaches are faster than standard training heuristics for the non-convex problem, such as SGD, and outperform commercial interior-point solvers. Experimentally, we verify our theoretical results, explore the group-$\ell_1$ regularization path, and scale convex optimization for neural networks to image classification on MNIST and CIFAR-10.
title Fast Convex Optimization for Two-Layer ReLU Networks: Equivalent Model Classes and Cone Decompositions
topic Machine Learning
url https://arxiv.org/abs/2202.01331