[USAD] 논문 리뷰

“USAD: UnSupervised Anomaly Detection on Multivariate Time Series”

Unsupervised Anomaly Detection AE-based multivariate time series AD

autoencoder에 adversarial training을 접목하여 AE-based 및 GAN-based multivariate time series anomaly detection 모델의 한계점을 보완한 
논문 리뷰 : https://hwk0702.github.io/treatise%20review/2021/02/15/USAD/

- Autoencoder 구조인 Encoder  E 와 Decoder D 를 이용
- original input vector  X 와 reconstruction  R 의 차이인 reconstruction error를 최소화
- 정상데이터의 reconstruction error를 기반으로 학습하여 정상 데이터의 분포를 학습함
- reconstruction error를 anomaly scores로 사용, threshold를 초과하면 이상치로 탐지함