def test_single_initializer_with_convo(initializer):
model = NeuralNetwork(
optimizer=AdamOptimizer(
learning_rate=default_parameters['learning_rate'] * 10),
loss=CrossEntropy(),
layers=[
Convolution2D(num_of_filters=8,
kernel=(3, 3),
activation_func=ReLu()),
MaxPooling2D(pool_size=(2, 2), stride=(2, 2)),
Flatten(),
Dense(layer_size=50,
activation_func=ReLu(),
weight_initializer=initializer),
Dense(layer_size=10,
activation_func=Softmax(),
weight_initializer=initializer)
],
callbacks=[
LoggerCallback(),
PlotCallback(f'./lab_3/initializers/{initializer.get_name()}')
])
model.fit(x_train=X_train,
y_train=y_train,
x_val=X_val,
y_val=y_val,
epochs=default_parameters['epochs'],
batch_size=default_parameters['batch_size'])
model.test(X_test, y_test)
def test_single_cost_and_last_layer(cost_func, last_layer):
model = NeuralNetwork(
optimizer=StaticGradientDescent(default_parameters['learning_rate']),
loss=cost_func,
layers=[
Dense(layer_size=50,
activation_func=ReLu(),
weight_initializer=XavierInitializer()),
Dense(layer_size=10,
activation_func=last_layer,
weight_initializer=XavierInitializer())
],
callbacks=[
LoggerCallback(),
PlotCallback(
f'./lab_3/cost/func={cost_func.get_name()}&last_layer={last_layer.get_name()}'
)
])
model.fit(x_train=X_train,
y_train=y_train,
x_val=X_val,
y_val=y_val,
epochs=default_parameters['epochs'],
batch_size=default_parameters['batch_size'])
model.test(X_test, y_test)
def test_single_initializer(initializer):
model = NeuralNetwork(
optimizer=AdamOptimizer(
learning_rate=default_parameters['learning_rate']),
loss=CrossEntropy(),
layers=[
Flatten(),
Dense(layer_size=50,
activation_func=ReLu(),
weight_initializer=initializer),
Dense(layer_size=10,
activation_func=Softmax(),
weight_initializer=initializer)
],
callbacks=[
LoggerCallback(),
PlotCallback(f'./lab_3/initializers/{initializer.get_name()}')
])
model.fit(x_train=X_train,
y_train=y_train,
x_val=X_val,
y_val=y_val,
epochs=default_parameters['epochs'],
batch_size=default_parameters['batch_size'])
model.test(X_test, y_test)
def __init__(self, num_of_filters, kernel, stride=(1, 1), kernel_initializer=XavierInitializer(),
bias_initializer=ZeroInitializer(),
activation_func=ReLu(), layer_name='Convo2D'):
super().__init__(layer_name)
self._num_of_filters = num_of_filters
self._kernel = kernel
self._stride = stride
self._kernel_initializer = kernel_initializer
self._bias_initializer = bias_initializer
self._activation_func = activation_func
self._z = None
def test_single_weight_initializer(weight_initializer):
model = NeuralNetwork(
optimizer=StaticGradientDescent(learning_rate=default_parameters['learning_rate']),
loss=CrossEntropy(),
layers=[
Dense(layer_size=50, activation_func=ReLu(), weight_initializer=weight_initializer),
Dense(layer_size=10, activation_func=Softmax(), weight_initializer=weight_initializer)
],
callbacks=[
LoggerCallback(),
PlotCallback(f'./results/weigh_initializer/{weight_initializer.get_name()}')
]
)
model.fit(x_train=X_train, y_train=y_train, x_val=X_val, y_val=y_val,
epochs=default_parameters['epochs'], batch_size=default_parameters['batch_size'])
model.test(X_test, y_test)
def single_batch_size_test(batch_size):
model = NeuralNetwork(
optimizer=StaticGradientDescent(learning_rate=default_parameters['learning_rate']),
loss=CrossEntropy(),
layers=[
Dense(layer_size=50, activation_func=ReLu(), weight_initializer=XavierInitializer()),
Dense(layer_size=10, activation_func=Softmax(), weight_initializer=XavierInitializer())
],
callbacks=[
LoggerCallback(),
PlotCallback('./results/batch_size/')
]
)
model.fit(x_train=X_train, y_train=y_train, x_val=X_val, y_val=y_val,
epochs=default_parameters['epochs'], batch_size=batch_size)
model.test(X_test, y_test)
# 'test_name': 'normal_C3x3-F1_MP2x2_F_D50_D10',
# 'layers': [
# Convolution2D(num_of_filters=1, kernel=(3, 3), activation_func=ReLu()),
# MaxPooling2D(pool_size=(2, 2), stride=(2, 2)),
# Flatten(),
# Dense(layer_size=50, activation_func=ReLu(), weight_initializer=HeInitializer()),
# Dense(layer_size=10, activation_func=Softmax(), weight_initializer=HeInitializer())
# ]
# },
{
'test_name':
'normal_C3x3-F2_MP2x2_F_D50_D10',
'layers': [
Convolution2D(num_of_filters=2,
kernel=(3, 3),
activation_func=ReLu()),
MaxPooling2D(pool_size=(2, 2), stride=(2, 2)),
Flatten(),
Dense(layer_size=50,
activation_func=ReLu(),
weight_initializer=HeInitializer()),
Dense(layer_size=10,
activation_func=Softmax(),
weight_initializer=HeInitializer())
]
},
{
'test_name':
'normal_C3x3-F4_MP2x2_F_D50_D10',
'layers': [
Convolution2D(num_of_filters=4,