y_test_log_svm[k][0] = i
x_test.append(test_temp[j].T)
k += 1
x_test = np.array(x_test)
x_test_vec = []
for i in range(len(x_test)):
x_temp = cv2.resize(x_test[i], (20, 17), interpolation=cv2.INTER_AREA)
x_test_vec.append(x_temp.flatten())
x_test_vec = np.array(x_test_vec)
# x_test_vec = ((x_test_vec - 128.0)/128.0) - 1
print(x_test_vec.shape)
test = PCA.PCA(d=340)
mean1, basis1, new_x_data_train = test.pca(x_train_vec.T)
mean2, basis2, new_x_data_test = test.pca(x_test_vec.T)
# print(new_x_data_train.shape)
test1 = NeuralNetwork.NeuralNet(100, 55, 10, actv='sigmoid')
test1.train(new_x_data_train, y_train, new_x_data_test, y_test)
lambda_set = [100, 0.01, 0.001, 10, 1, 0.1]
test1 = LogisticReg.MultiClassLog(20, 0.000005, lambda_set, 20000, 2)
param, x_test1 = test1.classification(new_x_data_train, new_x_data_test, y_test_log_svm)
test1.test(param, x_test1, y_test_log_svm)
test2 = SVM.MultiClassSVM(1000, 1, 0.001)
param = test2.classification(new_x_data_train, new_x_data_test, y_test_log_svm)
test2.test(param, new_x_data_test, y_test_log_svm)
