首页| English| 中国科学院

Anomaly Detection via Minimum Likelihood Generative Adversarial Networks

副标题:

时间:2018-11-05  来源:KLMM

题目:           Anomaly Detection via Minimum Likelihood Generative Adversarial Networks

报告人:      王础 (北京京航计算通讯研究所)

时间地点:   2018.11.15  10:00am  N204

摘要:           Anomaly detection aims to detect abnormal events by a model of normality. It plays an important role in many domains such as network intrusion detection,  criminal activity identity  and so on. With the rapidly growing size of accessible training data and high computation capacities, deep learning based anomaly detection has become more and more popular. In this paper, a new domain-based anomaly detection method based on generative adversarial networks (GAN) is proposed. Minimum likelihood regularization is proposed to make the  generator produce more anomalies and prevent it from converging to normal data distribution. Proper ensemble of anomaly scores is shown to improve the stability of discriminator effectively. The proposed method  has achieved significant improvement than other anomaly detection methods  on Cifar10 and UCI datasets.

相关附件
相关文档