D-Clustering: A Kernel Based Graph Clustering Algorithm

The weight of edges: 

               Page rank  -  The weight of the vertex divided by the forward link

               D-Clustering - The weight of the edge decided by the kernel function ( End vertex as input, edge weight as function output)          

The accuracy of the clustering result:

              The degree of the convergence determines the accuracy

Issue 1:

         Partial cluster overlap can emerge when shared cluster member is allowed.

      Solution:

         1. The liner search implementation is only suitable to problems when mutual class member is naturally unlikely to appear

         2. Force a strong kernel testing criteria