def __init__(self, latent_dim=100, sparsity_weight=1, sparsity_objective=0.1, decay_positive_weights=0, decay_negative_weights=1, decay_weight=1):
"""
Create a sparse shallow AE with the custom kl divergence regularizer, enforcing weights non negativity with an asymmetric decay.
Arguments:
sparsity_weight: positive float - the weight of the sparsity cost.
sparsity_objective: float between 0 and 1 - the sparsity parameter.
decay_positive_weights: positive float - the weight decay parameter for the positive weights.
decay_negative_weights: positive float - the weight decay parameter for the negative weights.
decay_weight: positive float - the weight of the whole non negativity cost.
"""
self.latent_dim = latent_dim
self.sparsity_weight = sparsity_weight
self.sparsity_objective = sparsity_objective
self.decay_positive_weights = decay_positive_weights
self.decay_negative_weights = decay_negative_weights
self.decay_weight = decay_weight
input_img = Input(shape=(28, 28, 1)) # adapt this if using `channels_first` image data format
x = Flatten()(input_img)
encoded = Dense(latent_dim, activation='sigmoid', activity_regularizer=custom_regularizers.KL_divergence(beta=self.sparsity_weight, rho=self.sparsity_objective), kernel_regularizer=custom_regularizers.asymmetric_weight_decay(alpha=self.decay_positive_weights, beta=self.decay_negative_weights, lam=self.decay_weight))(x)
self.encoder = Model(input_img, encoded, name='encoder')
encoded_img = Input(shape=(self.latent_dim,))
x = Dense(28*28)(encoded_img)
x = LeakyReLU(alpha=0.1)(x)
decoded = Reshape((28,28,1))(x)
self.decoder = Model(encoded_img, decoded, name='decoder')
encoded = self.encoder(input_img)
decoded = self.decoder(encoded)
self.autoencoder = Model(input_img, decoded)
self.autoencoder.compile(optimizer='adadelta', loss='mean_squared_error')
def __init__(self, latent_dim=100, beta=1, rho=0.1):
"""
Create a sparse shallow AE with the custom kl divergence regularizer.
Arguments:
beta: positive float - the weight of the sparsity cost.
rho: float between 0 and 1 - the sparsity parameter
"""
self.latent_dim = latent_dim
self.beta = beta
self.rho = rho
input_img = Input(shape=(
28, 28,
1)) # adapt this if using `channels_first` image data format
x = Flatten()(input_img)
encoded = Dense(latent_dim,
activation='sigmoid',
activity_regularizer=custom_regularizers.KL_divergence(
beta=self.beta, rho=self.rho))(x)
self.encoder = Model(input_img, encoded, name='encoder')
encoded_img = Input(shape=(self.latent_dim, ))
x = Dense(28 * 28)(encoded_img)
x = LeakyReLU(alpha=0.1)(x)
decoded = Reshape((28, 28, 1))(x)
self.decoder = Model(encoded_img, decoded, name='decoder')
encoded = self.encoder(input_img)
decoded = self.decoder(encoded)
self.autoencoder = Model(input_img, decoded)
self.autoencoder.compile(optimizer='adadelta',
loss='mean_squared_error')
def __init__(self, latent_dim=100, nb_rows=28, nb_columns=28, nb_input_channels=1, one_channel_output=True,
sparsity_weight=1, sparsity_objective=0.1):
"""
Create a sparse shallow AE with the custom kl divergence regularizer, enforcing weights non negativity with Keras NonNeg constraint.
Arguments:
sparsity_weight: positive float - the weight of the sparsity cost.
sparsity_objective: float between 0 and 1 - the sparsity parameter.
"""
self.latent_dim = latent_dim
self.nb_rows=nb_rows
self.nb_columns=nb_columns
self.nb_input_channels=nb_input_channels
if one_channel_output:
self.nb_output_channels=1
else:
self.nb_output_channels=nb_input_channels
self.sparsity_weight = sparsity_weight
self.sparsity_objective = sparsity_objective
input_img = Input(shape=(self.nb_rows, self.nb_columns, nb_input_channels)) # adapt this if using `channels_first` image data format
x = Flatten()(input_img)
encoded = Dense(latent_dim, activation='sigmoid',
activity_regularizer=custom_regularizers.KL_divergence(beta=self.sparsity_weight,
rho=self.sparsity_objective))(x)
self.encoder = Model(input_img, encoded, name='encoder')
encoded_img = Input(shape=(self.latent_dim,))
x = Dense(self.nb_rows*self.nb_columns*self.nb_output_channels,
kernel_constraint=constraints.non_neg())(encoded_img)
x = LeakyReLU(alpha=0.1)(x)
decoded = Reshape((self.nb_rows,self.nb_columns,self.nb_output_channels))(x)
self.decoder = Model(encoded_img, decoded, name='decoder')
encoded = self.encoder(input_img)
decoded = self.decoder(encoded)
self.autoencoder = Model(input_img, decoded)
self.autoencoder.compile(optimizer='adadelta', loss='mean_squared_error', metrics=['mse'])
def __init__(self, latent_dim=100, nb_rows=28, nb_columns=28, nb_input_channels=1, one_channel_output=True,
sparsity_weight=1, sparsity_objective=0.1):
"""
Create a sparse shallow AE with the custom kl divergence regularizer. Multiple losses are computed for each code coefficient (KL_div + rec_err)
"""
self.latent_dim = latent_dim
self.nb_input_channels=nb_input_channels
self.nb_rows=nb_rows
self.nb_columns=nb_columns
if one_channel_output:
self.nb_output_channels=1
else:
self.nb_output_channels=nb_input_channels
self.sparsity_weight = sparsity_weight
self.sparsity_objective = sparsity_objective
input_img = Input(shape=(self.nb_rows, self.nb_columns, self.nb_input_channels)) # adapt this if using `channels_first` image data format
x = Flatten()(input_img)
encoded = Dense(latent_dim, activation='sigmoid',
activity_regularizer=custom_regularizers.KL_divergence(beta=self.sparsity_weight,
rho=self.sparsity_objective))(x)
self.encoder = Model(input_img, encoded, name='encoder')
encoded_img = Input(shape=(self.latent_dim,))
x = Dense(self.nb_rows*self.nb_columns*self.nb_output_channels)(encoded_img)
x = LeakyReLU(alpha=0.1)(x)
decoded = Reshape((self.nb_rows,self.nb_columns,self.nb_output_channels))(x)
self.decoder = Model(encoded_img, decoded, name='decoder')
encoded = self.encoder(input_img)
decoded = self.decoder(encoded)
self.autoencoder = Model(input_img, decoded)
self.autoencoder.compile(optimizer='adadelta', loss='mean_squared_error', metrics=['mse'])
def createAndTrain(x_train,
y_train,
x_test,
y_test,
latent_dim,
strDate,
sparse=False,
save=False):
print('\n\n **************** Code size: ', latent_dim,
' **************** \n')
input_img = Input(
shape=(28, 28,
1)) # adapt this if using `channels_first` image data format
x = Flatten()(input_img)
if sparse:
encoded = Dense(
latent_dim,
activation='sigmoid',
activity_regularizer=custom_regularizers.KL_divergence())(x)
else:
encoded = Dense(latent_dim, activation='sigmoid')(x)
encoder = Model(input_img, encoded, name='encoder')
encoder.summary()
encoded_img = Input(
shape=(latent_dim,
)) # adapt this if using `channels_first` image data format
x = Dense(28 * 28)(encoded_img)
x = LeakyReLU(alpha=0.1)(x)
decoded = Reshape((28, 28, 1))(x)
decoder = Model(encoded_img, decoded, name='decoder')
encoded = encoder(input_img)
decoded = decoder(encoded)
autoencoder = Model(input_img, decoded)
autoencoder.compile(optimizer='adadelta', loss='mean_squared_error')
encoded = encoder(input_img)
decoded = decoder(encoded)
autoencoder = Model(input_img, decoded)
autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')
t0 = time.time()
autoencoder.fit(
x_train,
x_train,
epochs=200,
verbose=2,
batch_size=128,
shuffle=True,
validation_data=(x_test, x_test),
)
t1 = time.time()
if ((latent_dim % 50 == 0) or (save == True)):
model_path = dir + '_AEinfoGAN_' + str(latent_dim) + '.h5'
autoencoder.save(model_path)
training_time = t1 - t0
training_error = autoencoder.evaluate(x_train, x_train, verbose=0)
test_error = autoencoder.evaluate(x_test, x_test, verbose=0)
return training_time, training_error, test_error
