classifier.train()
w=kernel.get_subkernel_weights()
kernel.set_subkernel_weights(w)
# Plot ROC curve
subplot(111)
ROC_evaluation=ROCEvaluation()
ROC_evaluation.evaluate(classifier.apply(feats_train),Labels(trainlab))
roc = ROC_evaluation.get_ROC()
plot(roc[0], roc[1])
fill_between(roc[0],roc[1],0,alpha=0.1)
grid(True)
xlabel('FPR')
ylabel('TPR')
title('Train ROC (Width=%.3f, C1=%.3f, C2=%.3f) ROC curve = %.3f' % (10, classifier.get_C1(), classifier.get_C2(), ROC_evaluation.get_auROC()),size=10)
savefig("data/iri/mkl.png")
"""
subplot(222)
ROC_evaluation=ROCEvaluation()
ROC_evaluation.evaluate(classifier.apply(feats_test),Labels(testlab))
roc = ROC_evaluation.get_ROC()
plot(roc[0], roc[1])
fill_between(roc[0],roc[1],0,alpha=0.1)
grid(True)
xlabel('FPR')
ylabel('TPR')
title('Test ROC (Width=%.3f, C1=%.3f, C2=%.3f) ROC curve = %.3f' % (0, classifier.get_C1(), classifier.get_C2(), ROC_evaluation.get_auROC()),size=10)
print sign(out)
print testlab
# Plot ROC curve
figure()
ROC_evaluation=ROCEvaluation()
ROC_evaluation.evaluate(mkl.apply(),labels)
roc = ROC_evaluation.get_ROC()
print roc
plot(roc[0], roc[1])
fill_between(roc[0],roc[1],0,alpha=0.1)
text(mean(roc[0])/2,mean(roc[1])/2,'auROC = %.5f' % ROC_evaluation.get_auROC())
grid(True)
xlabel('FPR')
ylabel('TPR')
title('MKL PolyNomial Kernel (C=%.3f) ROC curve' % mkl.get_C1(),size=10)
figure()
plot(pos[0, :], pos[1, :], "g.")
plot(neg[0, :], neg[1, :], "r.")
grid(True)
title('Data',size=10)
x, y, z = compute_output_plot_isolines(mkl, kernel, feats_train)
pcolor(x, y, z, shading='interp')
contour(x, y, z, linewidths=1, colors='black', hold=True)
axis('tight')
show()
show()
