Semi-supervised learning using random subspace based linear embedding repulsion graph

Graph construction is a key step for graph based semi-supervised classification. Inspired the success of the local linear embedding method, this paper present a novel method called random subspace method based linear neighborhoods embedding repulsion graph (RSMLNER). In the proposed framework, the geodesic distance and input data labels are used to define neighborhood system, resulting that one can construct a good graph which can describe the data manifold more effectively. Moreover, the random subspace method is used for noise and redundancy suppression. Experiments on the Binary Alphadigits dataset, Newsgroups dataset and UCI dataset show the feasibility and effectiveness of the proposed method.