AAAI Publications, Twenty-Fourth International Joint Conference on Artificial Intelligence

Font Size: 
Multi-view Self-Paced Learning for Clustering
Chang Xu, Dacheng Tao, Chao Xu

Last modified: 2015-06-27


Exploiting the information from multiple views can improve clustering accuracy. However, most  existing multi-view clustering algorithms are non-convex and are thus prone to becoming stuck into bad local minima, especially when there are outliers and missing data. To overcome this problem, we present a new multi-view self-paced learning (MSPL) algorithm for clustering, that  learns the multi-view model by not only progressing from 'easy'  to 'complex' examples, but also from 'easy'  to 'complex' views. Instead of binarily separating the examples or views into 'easy' and 'complex', we design a novel probabilistic smoothed weighting scheme. Employing multiple views for clustering and  defining complexity  across both examples and views are shown theoretically  to be beneficial to optimal clustering. Experimental results on toy and real-world data demonstrate the efficacy of the proposed algorithm.

Full Text: PDF