Saved in:
Bibliographic Details
Main Authors: Wang, Ruigang, Dvijotham, Krishnamurthy, Manchester, Ian R.
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2402.01344
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929375732039680
author Wang, Ruigang
Dvijotham, Krishnamurthy
Manchester, Ian R.
author_facet Wang, Ruigang
Dvijotham, Krishnamurthy
Manchester, Ian R.
contents This paper presents a new bi-Lipschitz invertible neural network, the BiLipNet, which has the ability to smoothly control both its Lipschitzness (output sensitivity to input perturbations) and inverse Lipschitzness (input distinguishability from different outputs). The second main contribution is a new scalar-output network, the PLNet, which is a composition of a BiLipNet and a quadratic potential. We show that PLNet satisfies the Polyak-Lojasiewicz condition and can be applied to learn non-convex surrogate losses with a unique and efficiently-computable global minimum. The central technical element in these networks is a novel invertible residual layer with certified strong monotonicity and Lipschitzness, which we compose with orthogonal layers to build the BiLipNet. The certification of these properties is based on incremental quadratic constraints, resulting in much tighter bounds than can be achieved with spectral normalization. Moreover, we formulate the calculation of the inverse of a BiLipNet -- and hence the minimum of a PLNet -- as a series of three-operator splitting problems, for which fast algorithms can be applied.
format Preprint
id arxiv_https___arxiv_org_abs_2402_01344
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Monotone, Bi-Lipschitz, and Polyak-Lojasiewicz Networks
Wang, Ruigang
Dvijotham, Krishnamurthy
Manchester, Ian R.
Machine Learning
This paper presents a new bi-Lipschitz invertible neural network, the BiLipNet, which has the ability to smoothly control both its Lipschitzness (output sensitivity to input perturbations) and inverse Lipschitzness (input distinguishability from different outputs). The second main contribution is a new scalar-output network, the PLNet, which is a composition of a BiLipNet and a quadratic potential. We show that PLNet satisfies the Polyak-Lojasiewicz condition and can be applied to learn non-convex surrogate losses with a unique and efficiently-computable global minimum. The central technical element in these networks is a novel invertible residual layer with certified strong monotonicity and Lipschitzness, which we compose with orthogonal layers to build the BiLipNet. The certification of these properties is based on incremental quadratic constraints, resulting in much tighter bounds than can be achieved with spectral normalization. Moreover, we formulate the calculation of the inverse of a BiLipNet -- and hence the minimum of a PLNet -- as a series of three-operator splitting problems, for which fast algorithms can be applied.
title Monotone, Bi-Lipschitz, and Polyak-Lojasiewicz Networks
topic Machine Learning
url https://arxiv.org/abs/2402.01344