def predict_cnn(teX, teY):
_, _, h, w = teX.shape
start = time.process_time()
y_score = predict(teX)
end = time.process_time()
end = end-start
auc_r = auc.roc_auc(np.argmax(teY, axis=1), y_score)
return end, auc_r
w_h4 = init_weights((hidden, hidden))
w_o = init_weights((hidden, num_of_class)) # 10 <- output
noise_h, noise_h2, noise_h3, noise_h4, noise_py_x = model(X, w_h, w_h2, w_h3, w_h4, w_o, 0.2, 0.5)
h, h2, h3, h4, py_x = model(X, w_h, w_h2, w_h3, w_h4, w_o, 0., 0.)
y_x = T.argmax(py_x, axis=1)
cost = T.mean(T.nnet.categorical_crossentropy(noise_py_x, Y))
params = [w_h, w_h2, w_h3, w_h4, w_o]
updates = RMSprop(cost, params, lr=0.001)
train = theano.function(inputs=[X, Y], outputs=cost, updates=updates, allow_input_downcast=True)
predict = theano.function(inputs=[X], outputs=y_x, allow_input_downcast=True)
for i in range(100):
for start, end in zip(range(0, len(trX), 128), range(128, len(trX), 128)): # 0,128; 128,256; ...
cost = train(trX[start:end], trY[start:end])
y_score = predict(teX)
print "AUC:", auc.roc_auc(np.argmax(teY, axis=1), y_score)
deg, thresh, c1, c2, c3, c4, c5 = fusion_update(w_h, w_h2, w_h3, w_h4, w_o, rank)
y_score = predict(teX)
print "MF AUC:", auc.roc_auc(np.argmax(teY, axis=1), y_score), "deg:", np.mean(deg), "thresh:", np.mean(thresh)
for i in range(50):
for start, end in zip(range(0, len(trX), 128), range(128, len(trX), 128)):
cost = train(trX[start:end], trY[start:end])
repair(w_h, w_h2, w_h3, w_h4, w_o, c1, c2, c3, c4, c5)
y_score = predict(teX)
print "AUC:", auc.roc_auc(np.argmax(teY, axis=1), y_score)
