Clusters_search(γ): | |
---|---|
1. Clique and pseudoclique search: | • Find the maximal cliques and remove the cliques of size two if their nodes have more than 1 neighbour |
• For all cliques C, remove the nodes n for which max w(n, i)i∈C≤ τ(C) | |
• For all cliques C, add to C the γ-dense nodes n such that min w(n, i)i∈C≥ τ(C) | |
• Remove from the list of cluster the included clusters. | |
2. Select clusters according to their s-value: | • Remove the "worst clusters": For all clusters C1, C2 such that || ≥ ||, if | ∩ | ≥ γ × || and s(C1) ≥ s(C2) and Gelnb(C1) ≥ Gelnb(C2) then remove C2. |
• For all clusters C1, C2 such that | ∩ | ≥ γ × min(||, ||), add the nodes of min(s(C1), s(C2)) which have a maximal weight greater than τ(C min ) in max(s(C1), s(C2)). | |
3. Select nodes which belong to several clusters: | • For all nodes n which belong to several clusters, remove n from all clusters where it does not have its maximal MeanW. |