Re Learning Memory Guided Normality for Anomaly Detection

TL;DR

This paper proposes a novel unsupervised anomaly detection method using a Memory Module to enhance pattern learning and discriminative feature representation in neural networks, validated through visualizations and loss functions.

Contribution

Introduction of a Memory Module and two novel loss functions to improve unsupervised anomaly detection performance.

Findings

Memory Module improves pattern learning

Separateness and Compactness Losses increase discriminative power

t-SNE plots validate effective memory organization

Abstract

The authors have introduced a novel method for unsupervised anomaly detection that utilises a newly introduced Memory Module in their paper. We validate the authors claim that this helps improve performance by helping the network learn prototypical patterns, and uses the learnt memory to reduce the representation capacity of Convolutional Neural Networks. Further, we validate the efficacy of two losses introduced by the authors, Separateness Loss and Compactness Loss presented to increase the discriminative power of the memory items and the deeply learned features. We test the efficacy with the help of t-SNE plots of the memory items.