Journals
Resources
About Us
Open Access

Understanding the Initial Condensation of Convolutional Neural Networks

Understanding the Initial Condensation of Convolutional Neural Networks

Year:    2025

Author:    Zhangchen Zhou, Hanxu Zhou, Yuqing Li, Zhi-Qin John Xu

CSIAM Transactions on Applied Mathematics, Vol. 6 (2025), Iss. 2 : pp. 272–319

Abstract

Previous research has shown that fully-connected neural networks with small initialization and gradient-based training methods exhibit a phenomenon known as condensation [T. Luo et al., J. Mach. Learn. Res., 22(1), 2021]. Condensation is a phenomenon wherein the weight vectors of neural networks concentrate on isolated orientations during the training process, and it is a feature in the non-linear learning process that enables neural networks to possess better generalization abilities. However, the impact of neural network architecture on this phenomenon remains a topic of inquiry. In this study, we turn our focus towards convolutional neural networks (CNNs) to investigate how their structural characteristics, in contrast to fully-connected networks, exert influence on the condensation phenomenon. We first demonstrate in theory that under gradient descent and the small initialization scheme, the convolutional kernels of a two-layer CNN condense towards a specific direction determined by the training samples within a given time period. Subsequently, we conduct systematic empirical investigations to substantiate our theory. Moreover, our empirical study showcases the persistence of condensation under broader conditions than those imposed in our theory. These insights collectively contribute to advancing our comprehension of the non-linear training dynamics inherent in CNNs.

You do not have full access to this article.

Already a Subscriber? Sign in as an individual or via your institution

Journal Article Details

Publisher Name:    Global Science Press

Language:    English

DOI:    https://doi.org/10.4208/csiam-am.SO-2024-0011

CSIAM Transactions on Applied Mathematics, Vol. 6 (2025), Iss. 2 : pp. 272–319

Published online:    2025-01

AMS Subject Headings:    Global Science Press

Copyright:    COPYRIGHT: © Global Science Press

Pages:    48

Keywords:    Convolutional neural network dynamical regime condensation.

Author Details

Zhangchen Zhou

Hanxu Zhou

Yuqing Li

Zhi-Qin John Xu