Graph-based semi-supervised learning has received considerable attention and research,which has a good foundation in graph theory and is easy to solve.The performance of existing methods highly depends on the input weight graph,which is related to the hyper-parameters,such as the number of neighbors and the weight function.In many cases,it’s difficult to select appropriate hyper-parameters.In addition,there are multiple weights graphs representations for the same data set in many real-world applications.A key challenge is how to obtain better performance for graph-based semi-supervised learning by introducing multiple graphs.In this paper,a novel graph-based semi-supervised learning framework,self-paced multi-graphs label propagation learning(SPLP),is proposed.In our framework,the original problem is converted into a multi-graphs integration by setting different initial parameters to reduce the sensitivity of the algorithm to the selection of hype-parameters.For multi-graphs integration,the SPLP algorithm treats each weight graph as a sample,according to the self-paced learning theory,the learning process is divided into multiple stages,first training "simple" samples,and then gradually adding "complex" samples.By such learning strategy,SPLP can adaptively update the weight coefficient corresponding to each graph and has good robustness against noisy or irrelevant graphs.We experiment with the algorithm proposed in this article on multiple datasets,including image datasets(MNIST,PIE),text datasets(Sector,Reuters),and network datasets(Cora,Citeseer).Experimental results show that our method achieved better performance than other mainstream graph-based semi-supervised learning algorithms. |