Orthogonal optimal reverse prediction for semi-supervised learning

作者：

Highlights：

• We imposed the column orthogonal constraints to the mapping operator and propose the orthogonal optimal reverse prediction algorithm which shows a better performance than the original algorithm.

• We imposed the orthogonal constraints method to the kernelized optimal reverse prediction algorithm and proposed the kernelized orthogonal optimal reverse prediction.

• An optimization algorithm is designed to solve the problem and the model׳s classification performance is encouraging.

• We also extend the idea to the Laplacian orthogonal optimal reverse prediction and get a better classification performance.

摘要

Highlights•We imposed the column orthogonal constraints to the mapping operator and propose the orthogonal optimal reverse prediction algorithm which shows a better performance than the original algorithm.•We imposed the orthogonal constraints method to the kernelized optimal reverse prediction algorithm and proposed the kernelized orthogonal optimal reverse prediction.•An optimization algorithm is designed to solve the problem and the model׳s classification performance is encouraging.•We also extend the idea to the Laplacian orthogonal optimal reverse prediction and get a better classification performance.

论文关键词：Semi-supervised learning,Optimal reverse prediction,Orthogonal constraint,Orthogonal optimal reverse prediction

论文评审过程：Received 12 December 2015, Revised 1 July 2016, Accepted 3 July 2016, Available online 5 July 2016, Version of Record 26 July 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.07.005