adaboost_discriminant.m

function [c] = adaboost_discriminant( data, mu, sigma, p, alpha, classes, T)
%DISCRIMINANT Computes the maximum A Posteriori (MAP) given the data
%   INPUT data = data set, a MxN matrix
%   INPUT mu = mean for data (bayes_weight.m)
%   INPUT sigma = deviation for data (bayes_weight.m)
%   INPUT p = prior probability (prior.m using weights)
%   INPUT alpha =
%   INPUT classes =
%   INPUT t =
%   RETURN c = a mx1 vector
%
%   The discriminant function computes that given a pixel, how likely is it
%   to belong to the hand image or the hand plus book image

% data - MxN matrix
% mu   - CxN
% sigma- CxN
% g    - MxC
% c    - Mx1

N = size(data, 2);  % N - antal dimensioner
M = size(data,1);   % M - antal datapunkter
C = size(sigma, 1); % C - antal klasser

tempC = zeros(M,C);

for i = 1:T
    g  = discriminant(data, mu(:,:,i), sigma(:,:,i), p(i,:));
    [~, class] = max(g, [], 2);
    class = class - 1;
    
    for j = 1:C
        tempC(:,j) = tempC(:,j) + alpha(i) .* (class == classes(j));
    end
end

[~, c] = max(tempC, [], 2);
c = c - 1;