Deep Anomaly Detection by Residual Adaptation
arXiv:2010.02310
Abstract
Deep anomaly detection is a difficult task since, in high dimensions, it is hard to completely characterize a notion of "differentness" when given only examples of normality. In this paper we propose a novel approach to deep anomaly detection based on augmenting large pretrained networks with residual corrections that adjusts them to the task of anomaly detection. Our method gives rise to a highly parameter-efficient learning mechanism, enhances disentanglement of representations in the pretrained model, and outperforms all existing anomaly detection methods including other baselines utilizing pretrained networks. On the CIFAR-10 one-versus-rest benchmark, for example, our technique raises the state of the art from 96.1 to 99.0 mean AUC.
References in corpus (6)
- Deep Convolutional Inverse Graphics Network
- Deep Anomaly Detection with Outlier Exposure
- Multi-Level Variational Autoencoder: Learning Disentangled Representations from Grouped Observations
- BERT and PALs: Projected Attention Layers for Efficient Adaptation in Multi-Task Learning
- Deep Nearest Neighbor Anomaly Detection
- Latent Domain Learning with Dynamic Residual Adapters