svm

clc;
clear all;
close all;

load featureCheck.mat
load featureTrain.mat

featTrain=abs(featTrain);
featCheck=abs(featCheck);
[featTrain_norm, mu, sigma] = featureNormalize(featTrain);

%  Run PCA
[U, S] = pca(featTrain_norm);

tr=trace(S);
for i=1:size(S,1)
    br=trace(S(1:i,1:i));
    temp=br/tr;
    if(temp>=0.99)
        K=i;
        break
    end
end

%K = 5;
Z = projectData(featTrain_norm, U, K);
loss=1;
count=0;
% C=[0.01 0.03 0.1 0.3 1 3 10 30];
% mu=C;
C=6.2:0.05:6.6;
mu=4:0.05:4.4;

for i=1:length(C)
    for j=1:length(mu)
        t=templateSVM('boxConstraint',C(i),'KernelScale',mu(j),'Standardize',1,'KernelFunction','gaussian');
        Mdl=fitcecoc(Z,y_true_train,'Learners',t);
        CVMdl = crossval(Mdl);
        oosLoss = kfoldLoss(CVMdl);
        if(loss>oosLoss)
            loss=oosLoss;
            desiredC=C(i);
            desiredMu=mu(j);
        end
        count=count+1
    end
end

t=templateSVM('boxConstraint',desiredC,'KernelScale',desiredMu,'Standardize',1,'KernelFunction','gaussian');
Mdl=fitcecoc(Z,y_true_train,'Learners',t);