BriefGPT.xyz
May, 2024
基于边界的多类别泛化界限与几何复杂度
A Margin-based Multiclass Generalization Bound via Geometric Complexity
HTML
PDF
Michael Munn, Benoit Dherin, Javier Gonzalvo
TL;DR
通过对深度神经网络的一种复杂性度量，即几何复杂性，进行研究，我们提出了一种新的上界推导出的泛化误差，该泛化误差与网络的几何复杂性的边际归一化相关，并适用于广泛的数据分布和模型类。同时，我们对ResNet-18模型在CIFAR-10和CIFAR-100数据集上进行实验证明该广义化界是准确的。
Abstract
There has been considerable effort to better understand the
generalization capabilities
of
deep neural networks
both as a means to unlock a theoretical understanding of their success as well as providing directio
→