PARTIALLY SUPERVISED MACHINE LEARNING OF DATA CLASSIFICATION BASED ON LOCAL-NEIGHBORHOOD LAPLACIAN EIGENMAPS

摘要

A local-neighborhood Laplacian Eigenmap (LNLE) is provided for methods and systems for semi-supervised learning on manifolds of data points in a high-dimensional space. A labeled set and unlabeled data points are received as seen in Figure 4 (402). An adjacency Matrix/Graph is built (404). An unlabeled point is selected (406), then a local neighborhood/subgraph is found (408). Next, a Local Eigen Decomposition is computed (41) and evaluated (412) and the point is classified (414). A check is made to see if more points are available (416). If more points are available, select an unlabeled point (4Q6), otherwise output the classification (418).