#subtitleTextBox Clustering and the Voronoi Diagram
#mainTextBox Now we have added a DLL component which is similar to the discriminant function. It is called a Voronoi diagram and shows which regions of the input are clustered together. In our case since each cluster is defined by a PE, the Voronoi diagram displays the PE which will be the winner for each region of the input. The regions are color coded such that PE1 is white and the others are increasing shades of grey. Run the network and watch the Voronoi diagram change over time. Notice how the clustering has divided the input into three distinct clusters of data. This can then be used for classification.
#subtitleTextBox Another Input Set
#mainTextBox We have now created a new input data set which resembles six spokes on a wheel. Since there are six natural clusters of data, we have increased the number of PEs to 6. Run the network and watch as the PEs competitively try to "claim" their cluster. Watch the Voronoi diagram change over time as the PEs move throughout the input space.
#subtitleTextBox Summary
#mainTextBox In this example we showed the clustering which can be done using competitive learning and also introduced the Voronoi diagram to illustrate the clustering effect.
