资源简介
目前社区发现算法中计算速度最快的算法,由Vincent D.Blondel等人在2008年提出,基于modularity optimization启发式算法,代码可直接使用,在Vincent D.Blondel个人官网上下载的
代码片段和文件信息
% Iplementation : Antoine Scherrer
% antoine.scherrer@ens-lyon.fr
% Apply clustering after :
% “Fast unfolding of community hierarchies in large networks“
% Vincent D. Blondel Jean-Loup Guillaume Renaud Lambiotte
% Etienne Lefebvre
% http://arxiv.org/abs/0803.0476
%
% NON ORIENTED VERSION USING SYMETRIC MATRIX A = M + M^t INSTEAD OF
% (POSSIBLY NON SYMETRIC) INPUT MATRIX M
%
% FULL MATLAB VERSION (SLOWER)
%
% Inputs :
% M : weight matrix (the matrix is symetrized with
% the sum of weights in both directions)
% s : 1 = Recursive computation
% : 0 = Just one level computation
% self : 1 = Use self weights
% 0 = Do not use self weights
% debug : 1 = outputs some debug messages
% verbose : 1 = outputs some messages
%
% Output :
% COMTY structure with the following information
% for each level i :
% COMTY.COM{i} : vector of community IDs (sorted by community sizes)
% COMTY.SIZE{i} : vector of community sizes
% COMTY.MOD(i) : modularity of clustering
% COMTY.Niter(i) : Number of iteration before convergence
%
function [COMTY ending] = cluster_jl(Msselfdebugverbose)
if nargin < 1
error(‘not enough argument‘);
end
if nargin < 2
s = 1;
end
if nargin < 3
self = 1;
end
if nargin < 4
debug = 0;
end
if nargin < 5
verbose = 0;
end
S = size(M);
N = S(1);
ddebug = 0;
ending = 0;
% Symetrize matrix taking the sum of weights
M = M + M‘;
if (self == 0)
M((N+1).*[0:N-1]+1) = 0;
end
M2 = M;
M2((N+1).*[0:N-1]+1) = 0;
m = sum(sum(M));
Niter = 1;
if m==0 | N == 1
fprintf(‘No more possible decomposition\n‘);
ending = 1;
COMTY = 0;
return;
end
% Main loop
K = sum(M); % Sum of wieght incident to node i
SumTot = sum(M);
SumIn = diag(M); % Sum of weight inside community i
COM = 1:S(1); % Community of node i
for k=1:N
Neighbor{k} = find(M2(k:));
end
sCost = 10;
gain = 1;
while (gain == 1)
Cost = zeros(1N);
gain = 0;
for i=1:N
Ci = COM(i);
NB = Neighbor{i};
G = zeros(1N); % Gain vector
best_increase = -1;
Cnew = Ci;
COM(i) = -1;
SumTot(Ci) = SumTot(Ci) - K(i);
CNj1 = find(COM==Ci);
SumIn(Ci) = SumIn(Ci) - 2*sum(M(iCNj1)) - M(ii);
for j=1:length(NB)
Cj = COM(NB(j));
if (G(Cj) == 0)
CNj = find(COM==Cj);
Ki_in = 2*sum(M(iCNj));
G(Cj) = Ki_in/m - 2*K(i)*SumTot(Cj)/(m*m);
if (ddebug)
fprintf(‘Gaim for comm %d => %g\n‘Cj-1G(Cj));
end
if G(Cj) > best_increase;
best_increase = G(Cj);
Cnew_t = Cj;
end
end
end
if best_increase > 0
Cnew = Cnew_t;
if (debug)
fprintf(‘Move %d => %d\n‘i-1Cnew-1);
end
Cost(i) = best_increase;
end
Ck = find(COM==Cnew);
SumIn(Cnew) = SumIn(Cnew) + 2*sum(M(iCk));
SumTot(Cnew) = SumTot(Cnew) + K(i);
COM(i) = Cnew;
if (Cnew ~= Ci)
gain = 1;
end
end
sCost = sum(Cost);
[C2 S2] = reindex_com(COM);
Nco = length(unique(COM));
Nco2 = length(S2(S2>1));
mod = co
属性 大小 日期 时间 名称
----------- --------- ---------- ----- ----
文件 187 2016-06-26 18:24 Community_BGLL_Louvain\A2.mat
文件 5658 2008-07-15 13:52 Community_BGLL_Louvain\cluster_jl.m
文件 4166 2008-07-15 13:52 Community_BGLL_Louvain\cluster_jl_cpp.m
文件 5904 2016-05-30 21:41 Community_BGLL_Louvain\cluster_jl_orient.m
文件 5621 2008-07-15 13:59 Community_BGLL_Louvain\cluster_jl_orientT.m
文件 4056 2008-07-15 13:59 Community_BGLL_Louvain\cluster_jl_orientT_cpp.m
文件 3859 2008-07-15 13:52 Community_BGLL_Louvain\cluster_jl_orient_cpp.m
文件 5546 2008-07-15 13:52 Community_BGLL_Louvain\jl_clust.cpp
文件 13241 2008-07-15 13:52 Community_BGLL_Louvain\jl_clust.mexglx
文件 5856 2008-07-15 13:55 Community_BGLL_Louvain\jl_clust_orient.cpp
文件 14264 2008-07-15 13:56 Community_BGLL_Louvain\jl_clust_orient.mexglx
文件 1843 2008-06-26 15:32 Community_BGLL_Louvain\jl_mnew.cpp
文件 8756 2008-07-15 13:52 Community_BGLL_Louvain\jl_mnew.mexglx
文件 1502 2016-05-13 18:19 Community_BGLL_Louvain\readme.txt
目录 0 2016-07-23 15:30 Community_BGLL_Louvain
----------- --------- ---------- ----- ----
80459 15
- 上一篇:LNA 低噪放设计
- 下一篇:opencv.rar
相关资源
- A Dynamic Model for Check-in Behavior in Micro
- Rapid detection of virulent factors of Aeromon
- The assembly of a novel enzyme biosensor for a
- Variation of weed community composition in lat
- vs2017_Community 官方器
- Detection Estimation and Modulation Theory Par
- Functional redundancy instead of species redun
- Early detection in urinary proteome for the ef
- Implication of urinary KIM-1 mRNA detection in
- 虚假新闻检测的自然语言处理研究综
-
ba
sed.Radar.Detection.Tracking.And.Classifi - jude-community5.5.2中文版 收藏版
- JUDE/Community
- Learning Rich Features for Image Manipulation
- Quickest Detection_H. Vincent Poor(2008)
- deep learning for anomaly detection a survey(深
- vs2017_community.zip
- ship detection
- 基于内容的视频镜头检测
-
Mimicking Very Efficient Network for ob
ject - CNM社团检测算法
评论
共有 条评论