HyperGraph-based Minimax Probability Machines for Semi-Supervised Learning

Semi-Supervised Learning (SSL) is a method that combines both supervised as well as unsupervised learning to effectively learn from the humongous unlabelled data. In this paper, we introduce a Semi-Supervised Minimax Probability Machine model based on the Hypergraph Laplacian, leveraging hypergraphs to capture higher-order correlations within the intrinsic geometry of the manifold, thereby enhancing the learning process. Extensive experimental evaluations on well-known multi-label datasets showcase the efficacy and validity of our proposed model.