JLapVis -- A Java visualizer applet for the RLSC and LapRLSC learning algorithms.
Author: Mike Rainey
You need a Java-enabled browser to view this.

Download this applet as a tarball.

Instructions: To use JLapVis, start clicking on the point window. The points are initially in Class 1, but other classes can be chosen by clicking "Choose Class". After plotting some points, choose the learning algorithm to generate the contour map. RLSC ignores the unlabeled points, but LapRLSC recognizes them in proportion to the gammaI variable. The gammaA variable affects both algorithms; in a sense, gammaA determines the importance of the labeled points. The "K-nn" and "Eps-nn" options choose either K-nearest neighbors or Epsilon neighbors for calculating the Laplacian. The values for k-nearest neighbors or radius are set at the bottom. They only matter for unlabeled points in LapRLSC. When the values are finally set, click the "Draw" button to display the contour map.

Description: This applet allows the user to plot classified points in 2-d, and generates a contour map of RLSC or LapRLSC classifying the viewable space. RLSC and LapRLSC are binary classifiers. However, LapRLSC can also learn from unclassified or unlabeled points -- thus the user can choose points from class 1, class 2, or unlabeled. A more detailed explanation of RLSC and LapRLSC is at the following addresses:
Manifold Regularization
Also, here's the journal paper:
Manifold Regularization: a Geometric Framework for Learning from Examples