gd_buildsymmetricknngraph.m

来自「一个Matlab写的关于图理论以及其在机器学习中应用的教学用GUI软件」· M 代码 · 共 32 行

function W = GD_BuildSymmetricKnnGraph(M,k,which_matrix)
% Usage:  W = GD_BuildSymmetricKnnGraph(M,k,which_matrix)
%
% Input: 
% M                = either the distance or the similarity matrix, needs to be square
% k                = parameter of the knn graph
% which_matrix     = either 'sim' or 'dist'
% 
% Output: 
% W              = adjacency matrix of the symmetric knn graph without self-edges
%                  
%
% For a similarity matrix S, returns the undirected knn graph, edges are weighted by S. 
% For a distance matrix D, returns the undirected (unweighted!) knn graph. If you want to get 
% a weighted graph in this case, you need to take care of transforming D to S yourself and then 
% call the function with a similarity matrix 
%
% Excludes self-edges. 

%implemented brute force 



if (size(M,1) ~= size(M,2))
  error('Matrix not square!')
end

% compute directed KNN graph: 
W = GD_BuildDirectedKnnGraph(M,k,which_matrix); 
% transform it to symmetric one: 
W = max(W,W');