Abstract

The key to the graph based semi-supervised learning algo- rithms for classification problems is how to construct the weight ma- trix of the p-nearest neighbor graph. A new method to construct the weight matrix is proposed and a graph based Subspace Semi-supervised Learning Framework (SSLF) is developed. The Framework aims to find an embedding transformation which respects the discriminant structure inferred from the labeled data, as well as the intrinsic geometrical struc- ture inferred from both the labeled and unlabeled data. By utilizing this framework as a tool, we drive three semi-supervised dimensional- ity reduction algorithms: Subspace Semi-supervised Linear Discriminant Analysis (SSLDA), Subspace Semi-supervised Locality Preserving Pro- jection (SSLPP), and Subspace Semi-supervised Marginal Fisher Analy- sis (SSMFA). The experimental results on face recognition demonstrate our subspace semi-supervised algorithms are able to use unlabeled sam- ples effectively.