One vs All SVM algorithm using Matlab

ovasvm.m

function accuracy = ovasvm(data, n)

data = shuffleData(data);
[train, test] = splitData(data);
[trainSet, valSet] = splitData(train);
[trainSet, valSet] = scaleData(trainSet, valSet);

A = 1:n;
C = [2^-10 2^-9 2^-8 2^-7 2^-6 2^-5 2^-4 2^-3 2^-2 2^-1 2^0 2^1 2^2 2^3 2^4 2^5 2^6 2^7 2^8 2^9 2^10];
accuracy = [];

for i = 1:length(C)
    options = svmlopt('C', C(i), 'Verbosity', 0);
    predict = [];
    
    for class = 1:n
        Model = ['Model', int2str(A(class)), 'VsAll'];
        x = invertData(trainSet, A(class));
        y = x(:, end);
        x(:, end) = [];
        
        svmlwrite('SVMLTrain', x, y);
        svm_learn(options, 'SVMLTrain', Model);
        clear SVMLTrain x y;
        
        ModelOutput = ['ModelOutput', int2str(A(class)), 'VsAll'];
        xVal = invertData(valSet, class);
        yVal = xVal(:, end);
        xVal(:, end) = [];
        
        svmlwrite('SVMLVal', xVal, yVal);
        
        svm_classify(options, 'SVMLVal', Model, ModelOutput);
        svmpredict = svmlread(ModelOutput);
        predict = [predict, svmpredict];
    end
    
    accuracy(i) = WinnerTakesAll(valSet, predict, A);
end

[elt, ind] = max(accuracy);
cOpt = C(ind);
display(cOpt);

% Testing using optimal C value

options = svmlopt('C', cOpt, 'Verbosity', 0);
predict = [];

for class = 1:n
    Model = ['Model', int2str(A(class)), 'VsAll'];
    x = invertData(train, A(class));
    y = x(:, end);
    x(:, end) = [];
    
    svmlwrite('SVMLTrain', x, y);
    svm_learn(options, 'SVMLTrain', Model);
    
    ModelOutput = ['Model', int2str(A(class)), 'VsAll']; 
    xTest = invertData(test, A(class));
    yTest = xTest(:, end);
    xTest(:, end) = [];
    
    svmlwrite('SVMLTest', xTest, yTest);
    
    svm_classify(options, 'SVMLTest', Model, ModelOutput);
    svmpredict = svmlread(ModelOutput);
    predict = [predict, svmpredict];
end

accuracy = WinnerTakesAll(test, predict, A);


display(accuracy);