Deep Anomaly Detection with Outlier Exposure
Dan Hendrycks University of California, Berkeley、Mantas Mazeika University of Chicago、Thomas Dietterich Oregon State University
Abstract
It is important to detect anomalous inputs when deploying machine learning systems. The use of larger and more complex inputs in deep learning magnifies the difficulty of distinguishing between anomalous and in-distribution examples. At the same time, diverse image and text data are available in enormous quantities. We propose leveraging these data to improve deep anomaly detection by training anomaly detectors against an auxiliary dataset of outliers, an approach we call Outlier Exposure (OE). This enables anomaly detectors to generalize and detect unseen anomalies. In extensive experiments on natural language processing and small- and large-scale vision tasks, we find that Outlier Exposure significantly improves detection performance. We also observe that cutting-edge generative models trained on CIFAR-10 may assign higher likelihoods to SVHN images than to CIFAR-10 images; we use OE to mitigate this issue. We also analyze the flexibility and robustness of Outlier Exposure, and identify characteristics of the auxiliary dataset that improve performance.
中文速览
深度学习模型在实际部署时经常遇到与训练数据分布不同的"异常输入"，而现有方法往往难以可靠地将其与正常输入区分开来。本文提出了一种名为"异常曝光"（Outlier Exposure，OE）的方法：在训练异常检测器时，额外引入一个与测试异常数据不重叠的辅助"异常数据集"，让模型主动学习区分正常与异常输入的启发式规则，从而具备泛化到训练中从未见过的异常类型的能力。研究者在自然语言处理和计算机视觉（包括小尺度和大尺度图像）的多项任务上进行了大量实验，发现OE能显著提升异常检测性能，并能修复顶尖生成模型对CIFAR-10类外数据（如SVHN）赋予比类内数据更高似然这一反直觉的失效现象。这项工作以极低的额外成本（只需借助互联网上海量的图像或文本数据）就能系统性地增强现有检测方法，对提升实际AI系统的安全性和可靠性具有重要的实用价值。
原文 arXiv:1812.04606;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1812.04606v3