[Paper Review] AER: Auto-Encoder with Regression for Time Series Anomaly Detection скачать в хорошем качестве

[Paper Review] AER: Auto-Encoder with Regression for Time Series Anomaly Detection

발표자: 석박통합과정 강형원 1. 논문 제목: AER: Auto-Encoder with Regression for Time Series Anomaly Detection (Wong, L., Liu, D., Berti-Equille, L., Alnegheimish, S., & Veeramachaneni, K., IEEE BigData 2022) 링크: https://arxiv.org/pdf/2212.13558.pdf 2. 논문 Overview Prediction, reconstruction-based를 융합한 새로운 아키텍처인 AER(Auto-encoder with Regression)을 제안 Smoothing function에서 생성된 sequence 시작에서 false positive를 줄이기 위해 anomaly score를 masking하는 아이디어를 제안 Prediction-based anomaly score를 정방향과 역방향으로 결합한 bi-directional anomaly score를 제안 3. keyword: Time series, Anomaly detection, Time series anomaly detection, AER, Auto-Encoder, Prediction-based, Reconstruction-based