def load_tf_vae(filepath: str, state_dict: Dict) -> tf.keras.Model:
"""
Load VAE.
Parameters
----------
filepath
Save directory.
state_dict
Dictionary containing the latent dimension and beta parameters.
Returns
-------
Loaded VAE.
"""
model_dir = os.path.join(filepath, 'model')
if not [f for f in os.listdir(model_dir) if not f.startswith('.')]:
logger.warning(
'No encoder, decoder or vae found in {}.'.format(model_dir))
return None
encoder_net = tf.keras.models.load_model(
os.path.join(model_dir, 'encoder_net.h5'))
decoder_net = tf.keras.models.load_model(
os.path.join(model_dir, 'decoder_net.h5'))
vae = VAE(encoder_net,
decoder_net,
state_dict['latent_dim'],
beta=state_dict['beta'])
vae.load_weights(os.path.join(model_dir, 'vae.ckpt'))
return vae
def __init__(self,
threshold: float = None,
score_type: str = 'mse', # TODO: reconstruction proba; make sure to infer correct distribution
vae: tf.keras.Model = None,
encoder_net: tf.keras.Sequential = None,
decoder_net: tf.keras.Sequential = None,
latent_dim: int = None,
samples: int = 10,
beta: float = 1.,
data_type: str = None
) -> None:
"""
VAE-based outlier detector.
Parameters
----------
threshold
Threshold used for outlier score to determine outliers.
score_type
Metric used for outlier scores. Either 'mse' (mean squared error) or
'proba' (reconstruction probabilities) supported.
vae
A trained tf.keras model if available.
encoder_net
Layers for the encoder wrapped in a tf.keras.Sequential class if no 'vae' is specified.
decoder_net
Layers for the decoder wrapped in a tf.keras.Sequential class if no 'vae' is specified.
latent_dim
Dimensionality of the latent space.
samples
Number of samples sampled to evaluate each instance.
beta
Beta parameter for KL-divergence loss term.
data_type
Optionally specify the data type (tabular, image or time-series). Added to metadata.
"""
super().__init__()
if threshold is None:
logger.warning('No threshold level set. Need to infer threshold using `infer_threshold`.')
self.threshold = threshold
self.score_type = score_type
self.samples = samples
# check if model can be loaded, otherwise initialize VAE model
if isinstance(vae, tf.keras.Model):
self.vae = vae
elif isinstance(encoder_net, tf.keras.Sequential) and isinstance(decoder_net, tf.keras.Sequential):
self.vae = VAE(encoder_net, decoder_net, latent_dim, beta=beta) # define VAE model
else:
raise TypeError('No valid format detected for `vae` (tf.keras.Model) '
'or `encoder_net` and `decoder_net` (tf.keras.Sequential).')
# set metadata
self.meta['detector_type'] = 'offline'
self.meta['data_type'] = data_type
latent_dim = 50
encoder_net = tf.keras.Sequential([
InputLayer(input_shape=(input_dim, )),
Dense(128, activation=tf.nn.relu),
Dense(latent_dim, activation=None)
])
decoder_net = tf.keras.Sequential([
InputLayer(input_shape=(latent_dim, )),
Dense(128, activation=tf.nn.relu),
Dense(input_dim, activation=tf.nn.sigmoid)
])
ae = AE(encoder_net, decoder_net)
vae = VAE(encoder_net, decoder_net, latent_dim)
tests = [ae, vae]
@pytest.fixture
def tf_v_ae_mnist(request):
# load and preprocess MNIST data
(X_train, _), (X_test, _) = tf.keras.datasets.mnist.load_data()
X = X_train.reshape(60000,
input_dim)[:1000] # only train on 1000 instances
X = X.astype(np.float32)
X /= 255
# init model, predict with untrained model, train and predict with trained model
model = request.param
X_recon_untrained = model(X).numpy()
def __init__(self,
threshold: float = None,
vae: tf.keras.Model = None,
model: tf.keras.Model = None,
encoder_net: tf.keras.Sequential = None,
decoder_net: tf.keras.Sequential = None,
latent_dim: int = None,
samples: int = 10,
beta: float = 0.,
data_type: str = None) -> None:
"""
VAE-based adversarial detector.
Parameters
----------
threshold
Threshold used for adversarial score to determine adversarial instances.
vae
A trained tf.keras model if available.
model
A trained tf.keras classification model.
encoder_net
Layers for the encoder wrapped in a tf.keras.Sequential class if no 'vae' is specified.
decoder_net
Layers for the decoder wrapped in a tf.keras.Sequential class if no 'vae' is specified.
latent_dim
Dimensionality of the latent space.
samples
Number of samples sampled to evaluate each instance.
beta
Beta parameter for KL-divergence loss term.
data_type
Optionally specifiy the data type (tabular, image or time-series). Added to metadata.
"""
super().__init__()
if threshold is None:
logger.warning(
'No threshold level set. Need to infer threshold using `infer_threshold`.'
)
self.threshold = threshold
self.samples = samples
self.model = model
for layer in self.model.layers: # freeze model layers
layer.trainable = False
# check if model can be loaded, otherwise initialize VAE model
if isinstance(vae, tf.keras.Model):
self.vae = vae
elif isinstance(encoder_net, tf.keras.Sequential) and isinstance(
decoder_net, tf.keras.Sequential):
self.vae = VAE(encoder_net, decoder_net, latent_dim,
beta=beta) # define VAE model
else:
raise TypeError(
'No valid format detected for `vae` (tf.keras.Model) '
'or `encoder_net` and `decoder_net` (tf.keras.Sequential).')
# set metadata
self.meta['detector_type'] = 'offline'
self.meta['data_type'] = data_type