One-class classification with Gaussian processes

作者：

Highlights：

• Novel one-class classifications methods derived from the Gaussian process framework.

• Various applications: visual object recognition, defect detection, bacteria recognition, attribute prediction, and background subtraction.

• Outperforms or achieves comparable performance to state-of-the-art approaches.

• In-depth analysis of hyperparameter influence and one-class classification aspects.

摘要

Highlights•Novel one-class classifications methods derived from the Gaussian process framework.•Various applications: visual object recognition, defect detection, bacteria recognition, attribute prediction, and background subtraction.•Outperforms or achieves comparable performance to state-of-the-art approaches.•In-depth analysis of hyperparameter influence and one-class classification aspects.

论文关键词：One-class classification,Novelty detection,Kernel methods,Gaussian processes,Visual object recognition

论文评审过程：Received 24 November 2012, Revised 7 May 2013, Accepted 1 June 2013, Available online 24 June 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.06.005