Semi-supervised logistic discrimination via graph-based regularization

We address the problem of constructing a nonlinear discriminant procedure based on both labeled and unlabeled data sets.Asemi-supervised logistic model with Gaussian basis functions is presented along with the technique of graph-based regularization. A crucial issue in modeling process is the choice of tuning parameters included in the nonlinear semisupervised logistic models. In order to select these adjusted parameters, we derive model selection criteria from the viewpoints of information theory and also the Bayesian approach. Some numerical examples are given to investigate the effectiveness of our proposed semisupervised modeling strategies.