BriefGPT.xyz
Feb, 2018
基于检索机制的卷积神经网络提高对抗样本的鲁棒性
Retrieval-Augmented Convolutional Neural Networks for Improved Robustness against Adversarial Examples
HTML
PDF
Jake Zhao, Kyunghyun Cho
TL;DR
本研究提出了一种检索增强的卷积网络，采用局部混合（local mixup）训练，旨在缓解异常对抗示例的影响，并改善拟合问题。在 CIFAR-10、SVHN 和 ImageNet 数据集上，通过对比实验证明所提出的方法在提高鲁棒性方面更好。
Abstract
We propose a
retrieval-augmented convolutional network
and propose to train it with
local mixup
, a novel variant of the recently proposed mixup algorithm. The proposed hybrid architecture combining a convolutiona
→