|
Abstract:
|
In classification, semisupervised learning occurs when a large amount of unlabeled data is available with only a small number of labeled data. In such a situation, the objective is to use unlabeled data to enhance predictability of classification. This talk presents a novel margin-based, semisupervised learning methodology, utilizing the grouping information from unlabeled data together with the concept of margins in a form of regularization controlling the interplay between labeled and unlabeled data. In addition, the generalization error is estimated using both labeled and unlabeled data for tuning. The methodology is implemented for support vector machines (SVM) as well as psi-learning through difference convex programming, which reduces to sequential quadratic programs. Our theoretical and numerical analyses suggest the proposed methodology achieved the desired objective.
|
