Multi-label learning with co-training based on semi-supervised regression
Meixiang Xu, Fuming Sun, Xiaojun Jiang · 2014
The goal of this paper is to categorize images with multiple labels based on semi-supervised learning. Conventional semi-supervised regression methods are predominantly used to solve single label problems. However, it is more common in many real-world practical applications that an instance can be associated with a set of labels simultaneously. In this paper, a novel multi-label learning method with co-training based on semi-supervised regression is proposed to process multi-label classifications. Experimental results on two real-world data sets demonstrate that the proposed method is applicable to multi-label learning problems and its effectiveness outperforms that of three exiting state-of-the-art algorithms.