|
Abstract:
|
Recently, a classification method that combines both classified and unclassified data, called semi-supervised learning, has received considerable attention in the statistical and machine learning literature. There have been various modeling approaches to exploit information from the sets of partially classified samples; e.g., a mixture modeling approach, a graphical modeling approach and so on. We propose a semi-supervised nonlinear logistic model based on Gaussian basis expansions with the help of regularization. A crucial issue in model building process is the choice of adjusted parameters included in the models. In order to select the adjusted parameters, we derive a Bayesian type criterion for evaluating models estimated by the method of regularization. Numerical examples are given to investigate the effectiveness of our semi-supervised discriminant procedure.
|
