kmeans_v1.m

Go to the documentation of this file.

function centers = kmeans_v1(initPosArray, symbVector, varargin)

% we want row vectors:
if size(initPosArray,2) == 1
    initPosArray = initPosArray.';
end

if size(symbVector,2) == 1
    symbVector = symbVector.';
end

% default values:
iterations = 10; % Max. number of iterations
tol = 2e-3;      % Convergence tolerance
gamma = 0.5;     % Adaptation step

% configure optional variables:
if nargin > 2
    for argidx = 1:2:nargin-2
        switch varargin{argidx}
            case 'gamma'
                gamma = varargin{argidx+1};
            case 'iterations'
                iterations = varargin{argidx+1};
            case 'tol'
                tol = varargin{argidx+1};
        end
    end
end

K_Centers = initPosArray;
deltaK    = initPosArray;
                                     
d = nan(length(K_Centers), length(symbVector)); % Alocate matrix of data-centers distance.

%% k-means
for n = 1:iterations
    
    for ii = 1:length(K_Centers) % Calculate distances from all datapoints to each center
         d(ii,:) = abs(K_Centers(ii)-symbVector);
    end
        [~, ind] = min(d);
        
    for ii = 1:length(K_Centers) 
        deltaK(ii)    = gamma*(K_Centers(ii)-mean(symbVector(ind == ii))); % Incremental position values of each center
        K_Centers(ii) = K_Centers(ii) - deltaK(ii);                        % Update center positions
    end
    
    if sum(abs(deltaK))/length(K_Centers) < tol % Convergence test (if average position increment is less than 2e-3, stop iterations)
        break;
    end
    
end
centers = K_Centers;
end