comparison toolboxes/distance_learning/mlr/distance/setDistanceDiag.m @ 0:e9a9cd732c1e tip

first hg version after svn

+function D = setDistanceDiag(X, W, Ifrom, Ito)
+%
+% D = setDistanceDiag(X, W, Ifrom, Ito)
+%
+%   X       = d-by-n data matrix
+%   W       = d-by-1 PSD matrix
+%   Ifrom   = k-by-1 vector of source points
+%   Ito     = j-by-1 vector of destination points
+%
+%   D = n-by-n matrix of squared euclidean distances from Ifrom to Ito
+%       D is sparse, and only the rows corresponding to Ifrom and
+%       columns corresponding to Ito are populated.
+
+    [d,n]       = size(X);
+    L           = W.^0.5;
+    
+    Vfrom       = bsxfun(@times, L, X(:,Ifrom));
+
+    if nargin == 4
+        Vto     = bsxfun(@times, L, X(:,Ito));
+    else
+        Vto     = bsxfun(@times, L, X);
+        Ito     = 1:n;
+    end
+
+    D = distToFrom(n, Vto, Vfrom, Ito, Ifrom);
+end