BriefGPT.xyz
Feb, 2024
深度学习中信息瓶颈的更严格界限
Tighter Bounds on the Information Bottleneck with Application to Deep Learning
HTML
PDF
Nir Weingarten, Zohar Yakhini, Moshe Butman, Ran Gilad-Bachrach
TL;DR
使用变分近似方法为信息瓶颈提供新的、更紧的下界，从而提高先前基于信息瓶颈的深度神经网络的性能，并显著增强分类深度神经网络的对抗鲁棒性。
Abstract
deep neural nets
(DNNs) learn
latent representations
induced by their downstream task, objective function, and other parameters. The quality of the learned representations impacts the DNN's generalization ability
→