GraphDensityCut
Graph Clustering with Density-Cut Junming Shao, Qinli Yang, Jinhu Liu and Stefan Kramerâ€
Understanding :
- Build a Density-connected tree (DCT) Density Connectivity Map: DCT characterizes the density connectivity of vertices in graphs in a local fashion. It is intuitive that similar vertices are densely connected together, and vice versa
- That tree is unique for each graph (see Theorem 1)
- Each element of the DCT represents a component
- We try to find the weakest edge in the DCT to create two partitions
- We remove all the edges in the original graph which define these two partitions
- The original graph know contains two partitions.
- We repeat the same process from step 1 on each partition.
- We stop when we are happy