def test_train():
from tensorflow.keras.datasets import mnist
np.random.seed(seed=1)
(X_train, y_train), (X_test, y_test) = mnist.load_data()
# Reshape and Normalize data
X_train = X_train.reshape(-1, 784).astype(np.float64) / 255.0 - 0.5
y_train = y_train.flatten().astype(np.int32)
input_dimension = X_train.shape[1]
indices = list(range(X_train.shape[0]))
# np.random.shuffle(indices)
number_of_samples_to_use = 500
X_train = X_train[indices[:number_of_samples_to_use]]
y_train = y_train[indices[:number_of_samples_to_use]]
multi_nn = MultiNN(input_dimension)
number_of_classes = 10
activations_list = [multi_nn.sigmoid, multi_nn.sigmoid, multi_nn.linear]
number_of_neurons_list = [50, 20, number_of_classes]
for layer_number in range(len(activations_list)):
multi_nn.add_layer(number_of_neurons_list[layer_number],
activation_function=activations_list[layer_number])
for layer_number in range(len(multi_nn.weights)):
W = multi_nn.get_weights_without_biases(layer_number)
np.random.seed(seed=1)
W = tf.Variable((np.random.randn(*W.shape) - 0.0) * 0.1,
trainable=True)
multi_nn.set_weights_without_biases(W, layer_number)
b = multi_nn.get_biases(layer_number=layer_number)
b = tf.Variable(np.zeros(b.shape) * 0, trainable=True)
multi_nn.set_biases(b, layer_number)
multi_nn.set_loss_function(multi_nn.cross_entropy_loss)
percent_error = []
for k in range(10):
multi_nn.train(X_train,
y_train,
batch_size=100,
num_epochs=20,
alpha=0.8)
percent_error.append(multi_nn.calculate_percent_error(
X_train, y_train))
confusion_matrix = multi_nn.calculate_confusion_matrix(X_train, y_train)
# print(np.array2string(np.array(percent_error), separator=","))
# print(np.array2string(np.array(confusion_matrix), separator=","))
assert np.allclose(percent_error, np.array( \
[0.488,0.208,0.102,0.02 ,0.002,0. ,0. ,0. ,0. ,0. ]), rtol=1e-3, atol=1e-3)
assert np.allclose(confusion_matrix, np.array( \
[[50., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 66., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 52., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 50., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 52., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 39., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 45., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 52., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 0., 39., 0.],
[0., 0., 0., 0., 0., 0., 0., 0., 0., 55.]]), rtol=1e-3, atol=1e-3)
def test_train():
from tensorflow.keras.datasets import mnist
np.random.seed(seed=1)
(X_train, y_train), (X_test, y_test) = mnist.load_data()
# Reshape and Normalize data
X_train = X_train.reshape(-1, 784).astype(np.float64) / 255.0 - 0.5
y_train = y_train.flatten().astype(np.int32)
X_test = X_test.reshape(-1, 784).astype(np.float64) / 255.0 - 0.5
y_test = y_test.flatten().astype(np.int32)
input_dimension = X_train.shape[1]
indices = list(range(X_train.shape[0]))
# np.random.shuffle(indices)
number_of_samples_to_use_for_training = 500
number_of_samples_to_use_for_testing = 100
X_train = X_train[indices[:number_of_samples_to_use_for_training]]
y_train = y_train[indices[:number_of_samples_to_use_for_training]]
X_test = X_test[indices[:number_of_samples_to_use_for_testing]]
y_test = y_test[indices[:number_of_samples_to_use_for_testing]]
multi_nn = MultiNN(input_dimension)
number_of_classes = 10
activations_list = ["Relu", "Relu", "Linear"]
number_of_neurons_list = [50, 20, number_of_classes]
for layer_number in range(len(activations_list)):
multi_nn.add_layer(number_of_neurons_list[layer_number],
transfer_function=activations_list[layer_number])
for layer_number in range(len(multi_nn.weights)):
W = multi_nn.get_weights_without_biases(layer_number)
W = tf.Variable((np.random.randn(*W.shape)) * 0.3, trainable=True)
multi_nn.set_weights_without_biases(W, layer_number)
b = multi_nn.get_biases(layer_number=layer_number)
b = tf.Variable(np.zeros(b.shape) * 0, trainable=True)
multi_nn.set_biases(b, layer_number)
confusion_matrix = multi_nn.calculate_confusion_matrix(X_train, y_train)
# print("************* Confusion Matrix with training data before training ***************\n", np.array2string(confusion_matrix, separator=","))
assert np.allclose(confusion_matrix, np.array( \
[[0., 0., 0., 0., 0., 0., 44., 3., 0., 3.],
[0., 0., 0., 0., 1., 0., 62., 2., 0., 1.],
[0., 0., 0., 0., 1., 0., 42., 3., 0., 6.],
[0., 0., 0., 0., 0., 0., 47., 1., 0., 2.],
[0., 0., 0., 0., 0., 0., 20., 16., 6., 10.],
[0., 0., 0., 0., 1., 0., 30., 5., 1., 2.],
[0., 0., 0., 0., 0., 0., 35., 4., 2., 4.],
[0., 0., 0., 0., 0., 0., 25., 12., 13., 2.],
[0., 0., 0., 0., 0., 0., 29., 1., 4., 5.],
[0., 0., 0., 0., 1., 0., 43., 3., 8., 0.]]), rtol=1e-3, atol=1e-3)
percent_error_with_training_data = []
percent_error_with_test_data = []
for k in range(10):
multi_nn.train(X_train,
y_train,
batch_size=100,
num_epochs=20,
alpha=0.1)
percent_error_with_training_data.append(
multi_nn.calculate_percent_error(X_train, y_train))
percent_error_with_test_data.append(
multi_nn.calculate_percent_error(X_test, y_test))
confusion_matrix = multi_nn.calculate_confusion_matrix(X_train, y_train)
# print("************* Percent error using train ***************\n",np.array2string(np.array(percent_error_with_training_data), separator=","))
# print("************* Confusion Matrix with training data ***************\n", np.array2string(confusion_matrix, separator=","))
assert np.allclose(percent_error_with_training_data, np.array( \
[0.324, 0.14, 0.084, 0.036, 0.022, 0.014, 0.012, 0.012, 0.012, 0.012]), rtol=1e-3, atol=1e-3)
assert np.allclose(confusion_matrix, np.array( \
[[50., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 65., 1., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 52., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 1., 48., 0., 1., 0., 0., 0., 0.],
[0., 0., 0., 0., 52., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 39., 0., 0., 0., 0.],
[0., 0., 2., 0., 0., 0., 43., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 52., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 0., 39., 0.],
[0., 0., 0., 0., 0., 1., 0., 0., 0., 54.]]), rtol=1e-3, atol=1e-3)
confusion_matrix = multi_nn.calculate_confusion_matrix(X_test, y_test)
# print("************* Percent error using test ***************\n",np.array2string(np.array(percent_error_with_test_data), separator=","))
# print("************* Confusion Matrix with test data ***************\n", np.array2string(confusion_matrix, separator=","))
assert np.allclose(percent_error_with_test_data, np.array( \
[0.51, 0.36, 0.3, 0.28, 0.28, 0.27, 0.26, 0.26, 0.26, 0.26]), rtol=1e-3, atol=1e-3)
assert np.allclose(confusion_matrix, np.array( \
[[7., 0., 0., 0., 0., 1., 0., 0., 0., 0.],
[0., 12., 0., 0., 0., 0., 0., 0., 0., 2.],
[0., 0., 4., 1., 1., 1., 0., 1., 0., 0.],
[0., 0., 0., 9., 0., 1., 0., 0., 0., 1.],
[1., 0., 0., 0., 11., 0., 0., 1., 0., 1.],
[0., 0., 0., 1., 2., 2., 0., 0., 1., 1.],
[0., 0., 1., 0., 1., 0., 7., 1., 0., 0.],
[0., 0., 0., 0., 1., 0., 0., 12., 0., 2.],
[0., 0., 0., 0., 0., 1., 0., 0., 1., 0.],
[0., 0., 0., 0., 1., 0., 0., 1., 0., 9.]]), rtol=1e-3, atol=1e-3)
