Mercurial > hg > camir-aes2014

annotate toolboxes/distance_learning/mlr/distance/setDistanceDODMKL.m @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
rev   line source
wolffd@0 1 function D = setDistanceDODMKL(X, W, Ifrom, Ito)
wolffd@0 2 %
wolffd@0 3 % D = setDistanceDODMKL(X, W, Ifrom, Ito)
wolffd@0 4 %
wolffd@0 5 % X = d-by-n data matrix
wolffd@0 6 % W = m-by-m-by-n PSD matrix
wolffd@0 7 % Ifrom = k-by-1 vector of source points
wolffd@0 8 % Ito = j-by-1 vector of destination points
wolffd@0 9 %
wolffd@0 10 % D = n-by-n matrix of squared euclidean distances from Ifrom to Ito
wolffd@0 11 % D is sparse, and only the rows corresponding to Ifrom and
wolffd@0 12 % columns corresponding to Ito are populated.
wolffd@0 13
wolffd@0 14 [d,n,m] = size(X);
wolffd@0 15 L = W.^0.5;
wolffd@0 16
wolffd@0 17 D = 0;
wolffd@0 18 for i = 1:m
wolffd@0 19 for j = i:m
wolffd@0 20 Vfrom = bsxfun(@times, squeeze(L(i,j)), X(:,Ifrom,i) + X(:,Ifrom,j));
wolffd@0 21
wolffd@0 22 if nargin == 4
wolffd@0 23 Vto = bsxfun(@times, squeeze(L(i,j)), X(:,Ito,i) + X(:,Ito,j));
wolffd@0 24 else
wolffd@0 25 Vto = bsxfun(@times, squeeze(L(i,j)), X(:,:,i) + X(:,:,j));
wolffd@0 26 Ito = 1:n;
wolffd@0 27 end
wolffd@0 28
wolffd@0 29 if i == j
wolffd@0 30 s = 0.5;
wolffd@0 31 else
wolffd@0 32 s = 1;
wolffd@0 33 end
wolffd@0 34
wolffd@0 35 D = D + s * distToFrom(n, Vto, Vfrom, Ito, Ifrom);
wolffd@0 36 end
wolffd@0 37 end
wolffd@0 38
wolffd@0 39 end