Optimization Approaches for Semi-Supervised Multiclass Classification
Yasutoshi Yajima, Tien-Fang Kuo · 2006
The purpose of this paper is to propose a semi-supervised learning method for the problem of multiclass classification. We first introduce the Laplacian of a graph and the associated graph kernels which are exploited in many semi-supervised binary classification methods. Then, we will introduce a new multiclass semi-supervised learning method based on a multiclass formulation of SVM. The proposed optimization problems can fully exploit the sparse structure of the Laplacian matrix, which enables us to optimize the problems with a large number of data points by standard optimization algorithms. Some numerical results indicate that our approaches achieve fairly high performance on large scale problems