def __init__(self, dataset, *args, **kwargs):
super(CNN, self).__init__(*args, **kwargs)
model = Sequential()
model.add(
Conv2D(32, (3, 3), padding='same',
input_shape=dataset.input_shape))
model.add(Activation('relu'))
model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(64, (3, 3), padding='same'))
model.add(Activation('relu'))
model.add(Conv2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(dataset.output_size))
model.add(Activation('softmax'))
# opt = optimizers.rmsprop(lr=0.0001, decay=1e-6)
# opt = optimizers.adam()
opt = optimizers.sgd(lr=0.0001)
model.compile(loss='categorical_crossentropy',
optimizer=opt,
metrics=['accuracy'])
self.model = model
probabilistic_model = Sequential()
probabilistic_model.add(
Conv2D(32, (3, 3), padding='same',
input_shape=dataset.input_shape))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(Conv2D(32, (3, 3)))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(MaxPooling2D(pool_size=(2, 2)))
probabilistic_model.add(BayesianDropout(0.25))
probabilistic_model.add(Conv2D(64, (3, 3), padding='same'))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(Conv2D(64, (3, 3)))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(MaxPooling2D(pool_size=(2, 2)))
probabilistic_model.add(BayesianDropout(0.25))
probabilistic_model.add(Flatten())
probabilistic_model.add(Dense(512))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(BayesianDropout(0.5))
probabilistic_model.add(Dense(dataset.output_size))
probabilistic_model.add(Activation('softmax'))
# opt = optimizers.rmsprop(lr=0.0001, decay=1e-6)
# opt = optimizers.adam()
opt = optimizers.sgd(lr=0.0001)
probabilistic_model.compile(loss='categorical_crossentropy',
optimizer=opt,
metrics=['accuracy'])
self.probabilistic_model = probabilistic_model
def __init__(self, dataset, *args, **kwargs):
super(VGGTOP, self).__init__(*args, **kwargs)
cfg = {
'VGG11': [64, 'M', 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
'VGG13': [64, 64, 'M', 128, 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
'VGG16': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M'],
'VGG19': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 256, 'M', 512, 512, 512, 512, 'M', 512, 512, 512, 512, 'M'],
}
weight_decay = 0.0005
# deterministic model
model = Sequential()
model.add(Conv2D(64, (3, 3), padding='same', input_shape=dataset.input_shape, kernel_regularizer=l2(weight_decay)))
for x in cfg['VGG16'][1:]:
if x == 'M':
model.add(MaxPooling2D(pool_size=(2, 2)))
else:
model.add(Conv2D(x, (3, 3), padding='same', kernel_regularizer=l2(weight_decay)))
model.add(Activation('relu'))
model.add(BatchNormalization())
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512, kernel_regularizer=l2(weight_decay)))
model.add(Activation('relu'))
model.add(BatchNormalization())
model.add(Dropout(0.5))
model.add(Dense(dataset.output_size))
model.add(Activation('softmax'))
opt = optimizers.Adam()
model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
self.model = model
# probabilistic model
probabilistic_model = Sequential()
probabilistic_model.add(Conv2D(64, (3, 3), padding='same', input_shape=dataset.input_shape, kernel_regularizer=l2(weight_decay)))
for x in cfg['VGG16'][1:]:
if x == 'M':
probabilistic_model.add(MaxPooling2D(pool_size=(2, 2)))
else:
probabilistic_model.add(Conv2D(x, (3, 3), padding='same', kernel_regularizer=l2(weight_decay)))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(BatchNormalization())
probabilistic_model.add(BayesianDropout(0.25))
probabilistic_model.add(Flatten())
probabilistic_model.add(Dense(512, kernel_regularizer=l2(weight_decay)))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(BatchNormalization())
probabilistic_model.add(BayesianDropout(0.5))
probabilistic_model.add(Dense(dataset.output_size))
probabilistic_model.add(Activation('softmax'))
opt = optimizers.Adam()
probabilistic_model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
self.probabilistic_model = probabilistic_model
def __init__(self, dataset, layers=[50], dropout=0.5, *args, **kwargs):
super(MLP, self).__init__(*args, **kwargs)
self.dropout = dropout
tau = 0.159707652696 # obtained from BO
lengthscale = 1e-2
reg = lengthscale**2 * (1 - dropout) / (2. * len(dataset.x_train) *
tau)
# deterministic model
model = Sequential()
# model.add(Dropout(dropout, input_shape=dataset.input_shape))
model.add(Dense(layers[0], input_shape=dataset.input_shape))
model.add(Dropout(dropout))
model.add(Activation('relu'))
for units in layers[1:]:
model.add(Dense(units))
model.add(Dropout(dropout))
model.add(Activation('relu'))
model.add(Dense(dataset.output_size))
# probabilistic model
probabilistic_model = Sequential()
# probabilistic_model.add(BayesianDropout(dropout, input_shape=dataset.input_shape))
probabilistic_model.add(
Dense(layers[0], input_shape=dataset.input_shape))
probabilistic_model.add(BayesianDropout(dropout))
probabilistic_model.add(Activation('relu'))
for units in layers[1:]:
probabilistic_model.add(Dense(units))
probabilistic_model.add(BayesianDropout(dropout))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(Dense(dataset.output_size))
opt = optimizers.Adam()
if dataset.type == 'classification':
model.add(Activation('softmax'))
probabilistic_model.add(Activation('softmax'))
compile_params = {
'loss': 'categorical_crossentropy',
'optimizer': opt,
'metrics': ['accuracy']
}
else:
compile_params = {
'loss': nll_gaussian,
'optimizer': opt,
}
model.compile(**compile_params)
probabilistic_model.compile(**compile_params)
self.model = model
self.probabilistic_model = probabilistic_model
def __init__(self, input_shape, num_classes, epochs=10, batch_size=32):
self.input_shape = input_shape
self.num_classes = num_classes
self.epochs = epochs
self.batch_size = batch_size
model = Sequential()
model.add(Conv2D(32, (3, 3), padding='same', input_shape=input_shape))
model.add(Activation('relu'))
model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(64, (3, 3), padding='same'))
model.add(Activation('relu'))
model.add(Conv2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes))
model.add(Activation('softmax'))
opt = optimizers.rmsprop(lr=0.0001, decay=1e-6)
model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
self.model = model
probabilistic_model = Sequential()
probabilistic_model.add(Conv2D(32, (3, 3), padding='same', input_shape=input_shape))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(Conv2D(32, (3, 3)))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(MaxPooling2D(pool_size=(2, 2)))
probabilistic_model.add(BayesianDropout(0.25))
probabilistic_model.add(Conv2D(64, (3, 3), padding='same'))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(Conv2D(64, (3, 3)))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(MaxPooling2D(pool_size=(2, 2)))
probabilistic_model.add(BayesianDropout(0.25))
probabilistic_model.add(Flatten())
probabilistic_model.add(Dense(512))
probabilistic_model.add(Activation('relu'))
probabilistic_model.add(BayesianDropout(0.5))
probabilistic_model.add(Dense(num_classes))
probabilistic_model.add(Activation('softmax'))
opt = optimizers.rmsprop(lr=0.0001, decay=1e-6)
probabilistic_model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
self.probabilistic_model = probabilistic_model
def __init__(self, dataset, *args, **kwargs):
super(VGG, self).__init__(*args, **kwargs)
model = VGG16(include_top=False, input_shape=dataset.input_shape)
x = Flatten(name='flatten')(model.output)
x = Dense(512, activation='relu', name='fc1')(x)
x = Dropout(0.5)(x)
# x = Dense(512, activation='relu', name='fc2')(x)
# x = Dropout(0.5)(x)
x = Dense(dataset.output_size,
activation='softmax',
name='predictions')(x)
self.model = Model(inputs=model.input, outputs=x)
opt = optimizers.sgd(lr=0.0001)
self.model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
probabilistic_model = VGG16(include_top=False,
input_shape=dataset.input_shape)
x = Flatten(name='flatten')(probabilistic_model.output)
x = Dense(512, activation='relu', name='fc1')(x)
x = BayesianDropout(0.5)(x)
# x = Dense(512, activation='relu', name='fc2')(x)
# x = BayesianDropout(0.5)(x)
x = Dense(dataset.output_size,
activation='softmax',
name='predictions')(x)
self.probabilistic_model = Model(inputs=probabilistic_model.input,
outputs=x)
opt = optimizers.sgd(lr=0.0001)
self.probabilistic_model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])