%DEMO_ML_LFR - Finds communities of multilayer LFR benchmark using samSC.

%Performance of samSC is compared to regular SC on 40 realization of the

%benchmark. Multilayer LFR benchmark constructed with 5 layers, 200 nodes at

%each layer. Mixing coefficient is 0.5 and probability of within-layer

%communities forming an across-layer community is 0.5. For the remaining details

%about how the benchmark networks are constructed see [1].

%

% Other m-files required: sam_sc.m, spectral_clustering.m, calc_nmi.m

% MAT-files required: ml_lfr_mu-0.5_alpha-0.5.mat

%

% References:

% [1] Karaaslanli, Abdullah, and Selin Aviyente. "Strength Adjusted

% Multilayer Spectral Clustering." 2019 IEEE 29th International

% Workshop on Machine Learning for Signal Processing (MLSP). IEEE, 2019.

% Author: Abdullah Karaaslanli

% Address: Michigan State University, ECE

% email: karaasl1@msu.edu

% Website: http://www.abdkarr.github.io

% Date: 30-Oct-2020; Last revision: 30-Oct-2020

%

% Copyright (c) 2020, Abdullah Karaaslanli

% All rights reserved.

clear;

% load simulated multilayer networks: adjacencies, n_comms, gt_comms

load('ml_lfr_mu-0.5_alpha-0.5.mat');

n_exp = length(adjacencies);

nmi_sam_sc = zeros(1, n_exp);

nmi_ncut = zeros(1, n_exp);

for e=1:n_exp

% find communities with samsc

c_sam_sc = sam_sc(adjacencies{e}, n_comms{e});

% find communities with regular spectral_clustering

c_ncut = spectral_clustering(normalized_laplacian(cell2mat(adjacencies{e})), ...

n_comms{e}, true);

% NMIs

nmi_sam_sc(e) = calc_nmi(c_sam_sc, gt_comms{e});

nmi_ncut(e) = calc_nmi(c_ncut, gt_comms{e});

end

fprintf('NMI value of communities found by samSC: %.3f(+-%.3f)\n', ...

mean(nmi_sam_sc), std(nmi_sam_sc));

fprintf('NMI value of communities found by regular SC: %.3f(+-%.3f)\n', ...

mean(nmi_ncut), std(nmi_ncut));