lda_classify.m

一个matlab的工具包,里面包括一些分类器 例如 KNN KMEAN SVM NETLAB 等等有很多.

function  [Y_compute, Y_prob] = LDA_classify(para, X_train, Y_train, X_test, Y_test, num_class)

Y_compute = zeros(size(Y_test)); 
Y_prob = zeros(size(Y_test));
if (isempty(X_train)),
   error('The training set is empty!\n');
end;

[class_set, num_class] = GetClassSet(Y_train);
if (nargin <= 5), num_class = 2; end;

p = str2num(char(ParseParameter(para, {'-RegFactor'; '-QDA'}, {'0.1'; '0'})));
RegFactor = p(1);
QDA = p(2);

num_feature = size(X_train, 2);
num_test = size(Y_test, 1);
sigma = (1 - RegFactor) * cov(X_train) + RegFactor * eye(num_feature);
inv_sigma = inv(sigma);

data_mean = zeros(num_class, num_feature);
num_data_class = zeros(1, num_class);
Y_distance_matrix = zeros(num_test, num_class); 
for i = 1:num_class    
    % Convert the binary labels into +/-1
    data =  X_train(Y_train == class_set(i), :);
    data_mean(i, :) = mean(data);
    num_data_class(i) = size(data, 1);
    if (QDA > 0),
        sigma = (1 - RegFactor) * cov(data) + RegFactor * eye(num_feature);
        inv_sigma = inv(sigma); 
        num_data_class(i) = num_data_class(i) / sqrt(det(sigma));
    end;
    
    % Calculate the distance
    data_distance = X_test - repmat(data_mean(i, :), num_test, 1);
    Y_distance_matrix(:, i) = sum((data_distance * inv_sigma) .* data_distance, 2);
end;

[Y_distance Index] = min(Y_distance_matrix, [], 2);
Y_compute = class_set(Index);
Y_prob_matrix = repmat(num_data_class, num_test, 1) .* exp(-0.5 * Y_distance_matrix); 
Y_prob = max(Y_prob_matrix, [], 2) ./ sum(Y_prob_matrix, 2);