AdaFlow: Domain-Adaptive Density Estimator with Application to Anomaly Detection and Unpaired Cross-Domain Translation

[llien30]

INFO

Author

Masataka Yamaguchi, Yuma Koizumi, Noboru Harada

Affiliation

NTT

Conference or Year

ICASSP2019

Link

• arXiv
• [official GitHub]()

Abstract

Solved unsupervised anomaly detection(UAD) by using density estimation → Previous methods are difficult to adapt to the change in the normal data's distribution

Proposed Ada-Flow

• A unified model of a Normalizing Flow and Adaptive Batch Normalizations
• Can be adapted to a new distribution by just conducting forward propagation once per sample
• Can be used on devices that have limited computational sample
• Confirmed effectiveness of this method through an anomaly detection in a sound task
• Also confirmed that our model can be used for the cross-domain translation problem through experiments on image datasets.

  Proposed Method

Evaluation

Contribution

Discussion, Future Work

Comment

Date

2020/08/03