Optimization-Induced Dynamics of Lipschitz Continuity in Neural Networks

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Luo, Róisín, McDermott, James, Gagné, Christian, Sun, Qiang, O'Riordan, Colm
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908652028297216
author Luo, Róisín
McDermott, James
Gagné, Christian
Sun, Qiang
O'Riordan, Colm
author_facet Luo, Róisín
McDermott, James
Gagné, Christian
Sun, Qiang
O'Riordan, Colm
contents Lipschitz continuity characterizes the worst-case sensitivity of neural networks to small input perturbations; yet its dynamics (i.e. temporal evolution) during training remains under-explored. We present a rigorous mathematical framework to model the temporal evolution of Lipschitz continuity during training with stochastic gradient descent (SGD). This framework leverages a system of stochastic differential equations (SDEs) to capture both deterministic and stochastic forces. Our theoretical analysis identifies three principal factors driving the evolution: (i) the projection of gradient flows, induced by the optimization dynamics, onto the operator-norm Jacobian of parameter matrices; (ii) the projection of gradient noise, arising from the randomness in mini-batch sampling, onto the operator-norm Jacobian; and (iii) the projection of the gradient noise onto the operator-norm Hessian of parameter matrices. Furthermore, our theoretical framework sheds light on such as how noisy supervision, parameter initialization, batch size, and mini-batch sampling trajectories, among other factors, shape the evolution of the Lipschitz continuity of neural networks. Our experimental results demonstrate strong agreement between the theoretical implications and the observed behaviors.
format Preprint
id arxiv_https___arxiv_org_abs_2506_18588
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Optimization-Induced Dynamics of Lipschitz Continuity in Neural Networks
Luo, Róisín
McDermott, James
Gagné, Christian
Sun, Qiang
O'Riordan, Colm
Machine Learning
Artificial Intelligence
Lipschitz continuity characterizes the worst-case sensitivity of neural networks to small input perturbations; yet its dynamics (i.e. temporal evolution) during training remains under-explored. We present a rigorous mathematical framework to model the temporal evolution of Lipschitz continuity during training with stochastic gradient descent (SGD). This framework leverages a system of stochastic differential equations (SDEs) to capture both deterministic and stochastic forces. Our theoretical analysis identifies three principal factors driving the evolution: (i) the projection of gradient flows, induced by the optimization dynamics, onto the operator-norm Jacobian of parameter matrices; (ii) the projection of gradient noise, arising from the randomness in mini-batch sampling, onto the operator-norm Jacobian; and (iii) the projection of the gradient noise onto the operator-norm Hessian of parameter matrices. Furthermore, our theoretical framework sheds light on such as how noisy supervision, parameter initialization, batch size, and mini-batch sampling trajectories, among other factors, shape the evolution of the Lipschitz continuity of neural networks. Our experimental results demonstrate strong agreement between the theoretical implications and the observed behaviors.
title Optimization-Induced Dynamics of Lipschitz Continuity in Neural Networks
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2506.18588