A novel ordinal learning strategy: Ordinal nearest-centroid projection

作者：

Highlights：

• Propose an Ordinal Nearest-Centroid Projection (OrNCP) for ordinal regression (OR).

• Relax OrNCP to quadratic programming that covers the KDLOR and MOR as special cases.

• Experimentally demonstrate the effectiveness and superiority of our strategy for OR.

• Study the influences of the form and the granularity of ordinal constraints on OR.

摘要

•Propose an Ordinal Nearest-Centroid Projection (OrNCP) for ordinal regression (OR).•Relax OrNCP to quadratic programming that covers the KDLOR and MOR as special cases.•Experimentally demonstrate the effectiveness and superiority of our strategy for OR.•Study the influences of the form and the granularity of ordinal constraints on OR.

论文关键词：Ordinal regression,Ordinal nearest-centroid projection,Combinatorial optimization,Quadratic programming

论文评审过程：Received 30 September 2014, Revised 29 May 2015, Accepted 29 July 2015, Available online 1 August 2015, Version of Record 11 September 2015.

论文官网地址：https://doi.org/10.1016/j.knosys.2015.07.037