ARCADe: A Rapid Continual Anomaly Detector (2020-08-10T00:00:00.000000Z)

TL;DR

This work proposes A Rapid Continual Anomaly Detector (ARCADe), an approach to train neural networks to be robust against the major challenges of this new learning problem, namely catastrophic forgetting and overfitting to the majority class.

Abstract

Although continual learning and anomaly detection have separately been well-studied in previous works, their intersection remains rather unexplored. The present work addresses a learning scenario where a model has to incrementally learn a sequence of anomaly detection tasks, i.e. tasks from which only examples from the normal (majority) class are available for training. We define this novel learning problem of continual anomaly detection (CAD) and formulate it as a meta-learning problem. Moreover, we propose A Rapid Continual Anomaly Detector (ARCADe), an approach to train neural networks to be robust against the major challenges of this new learning problem, namely catastrophic forgetting and overfitting to the majority class. The results of our experiments on three datasets show that, in the CAD problem setting, ARCADe substantially outperforms baselines from the continual learning and anomaly detection literature. Finally, we provide deeper insights into the learning strategy yielded by the proposed meta-learning algorithm.

Authors

Volker Tresp

5 papers

A. Frikha

2 papers

Denis Krompass

2 papers

TL;DR

Abstract

Authors

References53 items

Generative Adversarial Networks

Learning to Continually Learn

Few-Shot One-Class Classification via Meta-Learning

Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAML

Meta-learnt priors slow down catastrophic forgetting in neural networks

Task Agnostic Continual Learning via Meta Learning

Meta-Learning Representations for Continual Learning

Variational Prototype Replays for Continual Learning

Generalizing from a Few Examples

Few-shot Learning: A Survey

Meta-Learning With Differentiable Convex Optimization

Task-Free Continual Learning

Learning to Learn without Forgetting By Maximizing Transfer and Minimizing Interference

Deep One-Class Classification

Adversarial Meta-Learning

Meta-learning with differentiable closed-form solvers

Reptile: a Scalable Metalearning Algorithm

Adversarially Learned One-Class Classifier for Novelty Detection

Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence

Meta-Learning and Universality: Deep Representations and Gradient Descent can Approximate any Learning Algorithm

Abnormal event detection in videos using generative adversarial nets

Meta-SGD: Learning to Learn Quickly for Few Shot Learning

Gradient Episodic Memory for Continual Learning

Overcoming Catastrophic Forgetting by Incremental Moment Matching

A Systematic Study of Online Class Imbalance Learning With Concept Drift

Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery

Prototypical Networks for Few-shot Learning

Continual Learning Through Synaptic Intelligence

Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks

Overcoming catastrophic forgetting in neural networks

Optimization as a Model for Few-Shot Learning

High-dimensional and large-scale anomaly detection using a linear one-class SVM with deep learning

Transfer representation-learning for anomaly detection

Progressive Neural Networks

Matching Networks for One Shot Learning

Prioritized Experience Replay

Learning Deep Representations of Appearance and Motion for Anomalous Event Detection

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

One-class classification: taxonomy of study and review of techniques

Support Vector Data Description

Anomaly detection: A survey

Outlier Detection Using Replicator Neural Networks

Estimating the Support of a High-Dimensional Distribution

Catastrophic forgetting in connectionist networks

Multitask Learning

Outlier Detection with Autoencoder Ensembles

Variational Autoencoder based Anomaly Detection using Reconstruction Probability

Neuron Clustering for Mitigating Catastrophic Forgetting in Supervised and Reinforcement Learning

One shot learning of simple visual concepts

This PDF file includes: Materials and Methods

Reducing the Dimensionality of Data with Neural

One-class classifier networks for target recognition applications

Evolutionary principles in self-referential learning, or on learning how to learn: The meta-meta-. hook

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names