Improving semi-supervised learning through optimum connectivity
作者：
Highlights：
• A new algorithm for semi-supervised learning based on optimum-path forest.
• The algorithm provides significant improvements in accuracy and efficiency.
• Labels are propagated from labeled to unlabeled training samples with less errors.
• The novel classifier can be more accurate than other state-of-the-art methods.
• A fast and effective algorithm suitable for developing active learning methods.
摘要
Highlights•A new algorithm for semi-supervised learning based on optimum-path forest.•The algorithm provides significant improvements in accuracy and efficiency.•Labels are propagated from labeled to unlabeled training samples with less errors.•The novel classifier can be more accurate than other state-of-the-art methods.•A fast and effective algorithm suitable for developing active learning methods.
论文关键词：Semi-supervised learning,Optimum-path forest classifiers
论文评审过程：Received 6 July 2015, Revised 2 February 2016, Accepted 28 April 2016, Available online 20 May 2016, Version of Record 1 June 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.04.020
