Active anomaly detection based on deep one-class classification

Pattern Recognition Letters(2023)

引用 1|浏览40
暂无评分
摘要
•Query strategy, which selects the most anomalous samples, is a sub-optimal approach for deep one-class classification.•We tackle two essential problems of active learning for deep SVDD; query strategy and semi-supervised learning method.•We propose uncertainty sampling with an adaptive boundary without data-dependent hyper-parameters.•We also propose a combination of deep SVDD with Noise contrastive estimation to construct a discriminative representation.•Our methods result in a favorable performance against the competing methods on seven anomaly detection benchmark datasets.
更多
查看译文
关键词
Deep anomaly detection,One-class classification,Deep SVDD,Active learning,Noise-contrastive estimation
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要