Lightly trained support vector data description for novelty detection

作者：

Highlights：

• A novel low-complexity anomaly detection algorithm based on SVDD is proposed.

• It computes the pre-image of the 'agent of the center' using SVDD Primal formulation.

• An efficient gradient-descent algorithm called SPSA is used to solve the Primal SVDD.

摘要

•A novel low-complexity anomaly detection algorithm based on SVDD is proposed.•It computes the pre-image of the 'agent of the center' using SVDD Primal formulation.•An efficient gradient-descent algorithm called SPSA is used to solve the Primal SVDD.

论文关键词：SVDD,Outlier detection,One-class classification,Scaling

论文评审过程：Received 3 December 2015, Revised 10 April 2017, Accepted 5 May 2017, Available online 5 May 2017, Version of Record 17 May 2017.

论文官网地址：https://doi.org/10.1016/j.eswa.2017.05.007