arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2312.03386cs.LGstat.ML
收藏

关于神经网络雅可比正则化训练的无限宽度分析

An Infinite-Width Analysis on the Jacobian-Regularised Training of a Neural Network

KAIST(韩国科学技术院);School of Computing(计算机学院)

Taeyoung Kim, Hongseok Yang

AI总结:

本文将深度神经网络的无限宽度分析扩展至雅可比矩阵,证明MLP及其雅可比在初始化时联合收敛至高斯过程,并揭示鲁棒训练下的演化由神经正切核变体决定的线性常微分方程描述,为雅可比正则化提供了理论洞察。

摘要:

最近对深度神经网络在其无限宽度极限下的理论分析,加深了我们对这些网络的初始化、特征学习和训练的理解,并为寻找合适的超参数、学习网络权重和执行推理带来了新的实用技术。在本文中,我们通过证明这种无限宽度分析可以扩展到深度神经网络的雅可比矩阵,拓宽了这一研究方向。我们表明,当多层感知机(MLP)隐藏层的宽度趋于无穷大时,MLP及其初始化时的雅可比矩阵联合收敛到一个高斯过程(GP),并刻画了该GP。我们还证明,在无限宽度极限下,MLP在所谓的鲁棒训练(即对雅可比矩阵进行正则化的训练)下的演化,由一个线性一阶常微分方程描述,该方程由神经正切核的一个变体决定。我们通过实验证明了我们的理论主张与宽有限网络的相关性,并经验性地分析了核回归解的性质,以获得对雅可比正则化的洞察。

英文摘要:

The recent theoretical analysis of deep neural networks in their infinite-width limits has deepened our understanding of initialisation, feature learning, and training of those networks, and brought new practical techniques for finding appropriate hyperparameters, learning network weights, and performing inference. In this paper, we broaden this line of research by showing that this infinite-width analysis can be extended to the Jacobian of a deep neural network. We show that a multilayer perceptron (MLP) and its Jacobian at initialisation jointly converge to a Gaussian process (GP) as the widths of the MLP's hidden layers go to infinity and characterise this GP. We also prove that in the infinite-width limit, the evolution of the MLP under the so-called robust training (i.e., training with a regulariser on the Jacobian) is described by a linear first-order ordinary differential equation that is determined by a variant of the Neural Tangent Kernel. We experimentally show the relevance of our theoretical claims to wide finite networks, and empirically analyse the properties of kernel regression solution to obtain an insight into Jacobian regularisation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑