Rethinking Assumptions in Deep Anomaly Detection
arXiv:2006.00339
Abstract
Though anomaly detection (AD) can be viewed as a classification problem (nominal vs. anomalous) it is usually treated in an unsupervised manner since one typically does not have access to, or it is infeasible to utilize, a dataset that sufficiently characterizes what it means to be "anomalous." In this paper we present results demonstrating that this intuition surprisingly seems not to extend to deep AD on images. For a recent AD benchmark on ImageNet, classifiers trained to discern between normal samples and just a few (64) random natural images are able to outperform the current state of the art in deep AD. Experimentally we discover that the multiscale structure of image data makes example anomalies exceptionally informative.
17 pages; accepted at the ICML 2021 Workshop on Uncertainty & Robustness in Deep Learning; An extended Journal paper of this work has been published in Transactions on Machine Learning Research: arXiv:2205.11474
References in corpus (8)
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- Semi-Supervised Learning with Deep Generative Models
- Deep Anomaly Detection with Outlier Exposure
- Deep Anomaly Detection Using Geometric Transformations
- Generative Probabilistic Novelty Detection with Adversarial Autoencoders
- Classification-Based Anomaly Detection for General Data
- Anomaly Detection With Multiple-Hypotheses Predictions
- Identifying and Categorizing Anomalies in Retinal Imaging Data
Cited by in corpus (7)
- Explainable Deep One-Class Classification
- Understanding Anomaly Detection with Deep Invertible Networks through Hierarchies of Distributions and Features
- Anomalous Sound Detection as a Simple Binary Classification Problem with Careful Selection of Proxy Outlier Examples
- Understanding the Effect of Bias in Deep Anomaly Detection
- Deep Random Projection Outlyingness for Unsupervised Anomaly Detection
- Deep Anomaly Detection by Residual Adaptation
- Data augmentation and pre-trained networks for extremely low data regimes unsupervised visual inspection