Robust Semi-Supervised Learning on Multiple Networks with Noise.

Graph-regularized semi-supervised learning has been effectively used for classification when (i) data instances are connected through a graph, and (ii) labeled data is scarce. Leveraging multiple relations (or graphs) between the instances can improve the prediction performance, however noisy and/or irrelevant relations may deteriorate the performance. As a result, an effective weighing scheme needs to be put in place for robustness. In this paper, we propose iMUNE, a robust and effective approach for multi-relational graph-regularized semi-supervised classification, that is immune to noise. Under a convex formulation, we infer weights for the multiple graphs as well as a solution (i.e., labeling). We provide a careful analysis of the inferred weights, based on which we devise an algorithm that filters out irrelevant and noisy graphs and produces weights proportional to the informativeness of the remaining graphs. Moreover, iMUNE is linearly scalable w.r.t. the number of edges. Through extensive experiments on various real-world datasets, we show the effectiveness of our method, which yields superior results under different noise models, and under increasing number of noisy graphs and intensity of noise, as compared to a list of baselines and state-of-the-art approaches.