function [label, knnmat] = knnr(sampledata, testdata, k)
% KNNR	K-nearest neighbor rule for classification
%	LABEL = KNNR(SAMPLE, INPUT, K)
%	SAMPLE: data set (last column: desired class)
%	INPUT: input (feature) matrix, with each row being the values
%		of a specific feature
%	K: the "k" in k nearest neighbor
%	LABEL: KNNR output
%
%	Note that the output label of the SAMPLE is assumed to take
%	the values from 1 to n, where n is the number of classes.

if nargin < 3, k = 1;, end

feature_n = size(sampledata,2)-1;
sample_input = sampledata(:, 1:feature_n);
sample_output = sampledata(:, feature_n+1);
out_label = countele(sample_output);	% possible output labels
class_n = length(out_label);
test_n = size(testdata, 1);		% no. of test input vectors
input = testdata(:, 1:feature_n);	% strip out label info, if any

% Distance matrix between sample_input and input vector
distmat = vecdist(sample_input, input);
[junk, index] = sort(distmat);

% knnmat(i,j) = class of i-th nearest point of j-th input vector
% (The size of knnmat is k times test_n.)
knnmat = reshape(sample_output(index(1:k,:)), k, test_n);

% label_count(i,j) = count of class-i points within j-th input vector's
% neighborhood
label_count = zeros(class_n, test_n);
for i = 1:test_n,
	[sorted_element, element_count] = countele(knnmat(:,i));
	label_count(sorted_element, i) = element_count;
end

[junk, label] = max(label_count);
label = label';