wolffd@0: function [y, l] = knnfwd(net, x)
wolffd@0: %KNNFWD	Forward propagation through a K-nearest-neighbour classifier.
wolffd@0: %
wolffd@0: %	Description
wolffd@0: %	[Y, L] = KNNFWD(NET, X) takes a matrix X of input vectors (one vector
wolffd@0: %	per row)   and uses the K-nearest-neighbour rule on the training data
wolffd@0: %	contained in NET to  produce  a matrix Y of outputs and a matrix L of
wolffd@0: %	classification labels. The nearest neighbours are determined using
wolffd@0: %	Euclidean distance. The IJth entry of Y counts the number of
wolffd@0: %	occurrences that an example from class J is among the K closest
wolffd@0: %	training examples to example I from X. The matrix L contains the
wolffd@0: %	predicted class labels as an index 1..N, not as 1-of-N coding.
wolffd@0: %
wolffd@0: %	See also
wolffd@0: %	KMEANS, KNN
wolffd@0: %
wolffd@0: 
wolffd@0: %	Copyright (c) Ian T Nabney (1996-2001)
wolffd@0: 
wolffd@0: 
wolffd@0: errstring = consist(net, 'knn', x);
wolffd@0: if ~isempty(errstring)
wolffd@0:   error(errstring);
wolffd@0: end
wolffd@0: 
wolffd@0: ntest = size(x, 1);		              % Number of input vectors.
wolffd@0: nclass = size(net.tr_targets, 2);		% Number of classes.
wolffd@0: 
wolffd@0: % Compute matrix of squared distances between input vectors from the training 
wolffd@0: % and test sets.  The matrix distsq has dimensions (ntrain, ntest).
wolffd@0: 
wolffd@0: distsq = dist2(net.tr_in, x);
wolffd@0: 
wolffd@0: % Now sort the distances. This generates a matrix kind of the same 
wolffd@0: % dimensions as distsq, in which each column gives the indices of the
wolffd@0: % elements in the corresponding column of distsq in ascending order.
wolffd@0: 
wolffd@0: [vals, kind] = sort(distsq);
wolffd@0: y = zeros(ntest, nclass);
wolffd@0: 
wolffd@0: for k=1:net.k
wolffd@0:   % We now look at the predictions made by the Kth nearest neighbours alone,
wolffd@0:   % and represent this as a 1-of-N coded matrix, and then accumulate the 
wolffd@0:   % predictions so far.
wolffd@0: 
wolffd@0:   y = y + net.tr_targets(kind(k,:),:);
wolffd@0: 
wolffd@0: end
wolffd@0: 
wolffd@0: if nargout == 2
wolffd@0:   % Convert this set of outputs to labels, randomly breaking ties
wolffd@0:   [temp, l] = max((y + 0.1*rand(size(y))), [], 2);
wolffd@0: end