Algorithm 1. Initial Clustering.

1: procedure INITIALCLUSTER G = (V, E), ${\sigma }_1^2$, C a number of initial clusters)

2:  Compute a shortest path between any u, v ∈ V.

3:  Choose C nodes uniformly at random, assign one label to each.

⊳ Each node is a starting point for a cluster.

4:  while some nodes are still unlabeled do

5:   for every cluster do

6:    Find neighbors of nodes in the given cluster.

7:    if there are unlabeled neighbors or labeled neighbors with a worse distance to another

cluster then

⊳ Keep only the first condition for a fast run.

8:    Add the neighbor whose addition minimizes the within-cluster evaluation function ${\hat{H}}_{{\sigma }_I^2}$.