Distributionally Robust Multiclass Classification and Applications in Deep Image Classifiers

Distributionally Robust Multiclass Classification and Applications in Deep Image Classifiers
复制标题

分布式鲁棒多类分类及其在深度图像分类器中的应用

DOI:
10.1109/icassp49357.2023.10095775
复制
发表时间:
2023
期刊:
and Signal Processing (ICASSP
影响因子:
--
通讯作者:
Paschalidis, Ioannis Ch.
Paschalidis, Ioannis Ch.
中科院分区:
--
文献类型:
--
作者:
Chen, Ruidi;Hao, Boran;Paschalidis, Ioannis Ch.

文献摘要

参考文献

被引文献

相似文献

我们开发了一个多类逻辑回归(MLR)的分布鲁棒优化(DRO)公式,它可以容忍被异常值污染的数据。DRO框架使用一个概率模糊集,定义为接近Wasserstein度量意义上的训练集的经验分布的分布球。我们将DRO公式松弛为正则化学习问题,其正则化器是系数矩阵的范数。我们为模型的解决方案建立了样本外性能保证,提供了正则化器在控制预测误差中的作用的见解。我们将提出的方法应用于使基于深度视觉变换(ViT)的[1]图像分类器对随机攻击和对抗性攻击具有鲁棒性。具体来说,使用MNIST和CIFAR-10数据集,通过采用一种新的随机训练方法,我们证明与基线方法相比,测试错误率降低了83.5%,损失降低了91.3%。
We develop a Distributionally Robust Optimization (DRO) formulation for Multiclass Logistic Regression (MLR), which could tolerate data contaminated by outliers. The DRO framework uses a probabilistic ambiguity set defined as a ball of distributions that are close to the empirical distribution of the training set in the sense of the Wasserstein metric. We relax the DRO formulation into a regularized learning problem whose regularizer is a norm of the coefficient matrix. We establish out-of-sample performance guarantees for the solutions to our model, offering insights on the role of the regularizer in controlling the prediction error. We apply the proposed method in rendering deep Vision Transformer (ViT)-based [1] image classifiers robust to random and adversarial attacks. Specifically, using the MNIST and CIFAR-10 datasets, we demonstrate reductions in test error rate by up to 83.5% and loss by up to 91.3% compared with baseline methods, by adopting a novel random training method.
DOI: 10.48550/arxiv.1608.00853
发表时间: 2016
期刊: --
影响因子: --
作者:
Dziugaite G
通讯作者: Dziugaite G
DOI: --
发表时间: 2019
期刊: Advances in Neural Information Processing Systems 32 (NIPS 2019
影响因子: --
作者:
Blanchet, Jose and
通讯作者: Blanchet, Jose and
利用深度学习和海运物流大数据预测航运市场状况的研究
DOI: --
发表时间: 2019
期刊:
影响因子: --
作者:
和田 祐次郎;河原 大輝;濱田 邦裕
通讯作者: 濱田 邦裕