# Set Features for Anomaly Detection

**Authors:** Niv Cohen, Issar Tzachor, Yedid Hoshen  
**arXiv:** [2311.14773](https://arxiv.org/abs/2311.14773)  
**Date:** 2023-11-24  
**Summary page:** https://www.summarizepaper.com/en/paper/2311.14773v1/

## Key points

- The paper introduces a novel approach to anomaly detection using set features
- The method characterizes samples based on the distribution of elements, rather than identifying anomalous parts within the sample
- By focusing on overall element distribution, the approach offers a more robust means of detecting anomalies expressed through unique combinations of normal elements
- The proposed method outperforms previous state-of-the-art techniques in image-level logical anomaly detection and sequence-level time series anomaly detection

## Summary

The paper "Set Features for Anomaly Detection" by Niv Cohen, Issar Tzachor, and Yedid Hoshen introduces a novel approach to detecting anomalies in samples that exhibit unusual combinations of normal elements. This innovative method utilizes set features to characterize each sample based on the distribution of its elements, rather than solely identifying anomalous parts within the sample. By focusing on the overall distribution of elements, this approach offers a more robust and effective means of detecting anomalies expressed through unique combinations of normal elements. The proposed method outperforms previous state-of-the-art techniques in both image-level logical anomaly detection and sequence-level time series anomaly detection, demonstrating its potential to enhance anomaly detection capabilities across various domains and applications.