Clustering

Clustering module

class markovianSBM.Clustering.Clustering(n, K)

Bases: object

Class that performs the final rounding step on the rows of the matrix $hat{B}$ which is the optimal solution of the SDP relaxation of the K-means problem.

Kmedoids(barx, bary, C)

Kmedoid algorithm that performs a rounding step on the rows of $hat{B}$.

Parameters

• barx – Output of the method ‘solve_relaxed_LP’

• bary – Output of the method ‘solve_relaxed_LP’

• C – Output of the method ‘solve_relaxed_LP’

build_partition(clust)

Given a list clust that associates to each node its community, this methods builds the associated partition of the noes of the graph.

Parameters

clust – List of estimated clusters for the n nodes of the graph

find_permutation(true_partition, approx_partition)

Find the permutation between the names of the true communities and the ones estimated by our algorithm.

Parameters

• true_partition – True partition of the nodes in the graph according to their clusters

• approx_partition – Estimated partition of the nodes

solve_relaxed_LP(M)

Prelimanary to run the K-medoid algorithm.

Parameters

M – M is the output of the relaxed SDP problem.

markovianSBM.Clustering.add_liste(Level, Liste, Node, UP=True)

markovianSBM.Clustering.recu(node, level, graph, liste, node2ind, dejavu, dico_closest, nodes)