Algorithm 3 — Consensus Clustering

Input: ${\{{\mathbf{S}}^{-p}\in {[0,1]}^{n^{-}\times K}\}}_{p=1}^P$ (P clusterings from Algorithm 1), K (number of clusters)

Output: ${\mathbf{S}}^{-}\in {[0,1]}^{n^{-}\times K}$ (Final Clustering Assignment)

• Compute co-occurrence matrix A using Eq. A.3

• Spectral clustering on A:

• Compute Laplacian matrix $\mathbf{L}=\text{diag}\left({\displaystyle {\sum }_{l=1}^{n^{-}}{\mathbf{A}}_{i,l}}\right)-\mathbf{A}$

• Compute the K eigenvectors (v1, …, vK) that correspond to K smallest eigenvalues of L (λ1 ≤ … ≤ λK)

• S− ← K-means([v1 … vK])