The representation is built in interations, where every iteration moves the 2D projections in order to minimize the error between the apparent distances and the real distances.
java jsammon dataseta window will appear with the projection of the dataset allong the two dimensions with the greater normalized variance.
Each point assumes a color corrsponding to its class; by moving the pointer near to a 2D point, you'll be able to see, in the text area below the graphic display, the real N-D coordinates of that point.
TheUp, Down, Left, Right, Zoom In and Zoom Out buttons allow navigation in the 2D space; Autoscale rescales the drawing so that all points are displayed.
With the "Iterate" button, Sammon's algorithm is started for the number of iterations set in the text field below the button.
Sammon's "Magic Factor is fixed at 0.3.
Some example datasets are supplied: helix.data and iris.data are two datasets used in Sammon's article, while vehicle.data is a subset of the dataset for a real-world problem.