Current Result Document :
| 한글제목(Korean Title) |
시계열 이상치 탐지를 위한 Autoencoder ODEs |
| 영문제목(English Title) |
Autoencoder ODEs for Time Series Anomaly Detection |
| 저자(Author) |
김병현
나지혜
변영신
김세진
이재길
Byunghyun Kim
Jihye Na
Yeongsin Byeon
Sejin Kim
Jae-Gil Lee
|
| 원문수록처(Citation) |
VOL 49 NO. 02 PP. 0547 ~ 0549 (2022. 12) |
한글내용
(Korean Abstract) |
|
영문내용
(English Abstract) |
Neural Ordinary Differential Equations (Neural ODEs) have allowed the hidden states of some neural networks to be continuous. Latent ODE models, which are variational autoencoders with an ODE solver as the decoder, benefit from this continuous property by naturally imputing irregularly-sampled time series. We modify the latent ODE model into a new model called autoencoder ODE (AE-ODE) for anomaly detection on time series data. Experimental results on a private multivariate time series dataset reveal that AE-ODE can be a competitive model for anomaly detection. |
| 키워드(Keyword) |
|
| 파일첨부 |
PDF 다운로드
|
