Most of graph-based methods for semi-supervised learning are transductive, giving predictions for only the unlabeled data in the training set, and not for an arbitrary test point. SLC(Semi-supervised Local Linear Coordinate), which is based on LLC(Local Linear Coordinate) is present here as an inductive method. The mixture of factor analyzers is used to model the raw data set, and the label smoothness over the graph is enforced by local approximation. At last, smooth nonlinear projection is achieved by local affine transformation. Experiment shows the s...
