def test_aegmm(aegmm_params):
# OutlierAEGMM parameters
threshold, n_gmm, w_energy, threshold_perc, return_instance_score = aegmm_params
# define encoder, decoder and GMM density net
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)
])
gmm_density_net = tf.keras.Sequential([
InputLayer(input_shape=(latent_dim + 2, )),
Dense(10, activation=tf.nn.relu),
Dense(n_gmm, activation=tf.nn.softmax)
])
# init OutlierAEGMM
aegmm = OutlierAEGMM(threshold=threshold,
encoder_net=encoder_net,
decoder_net=decoder_net,
gmm_density_net=gmm_density_net,
n_gmm=n_gmm)
assert aegmm.threshold == threshold
assert aegmm.meta == {
'name': 'OutlierAEGMM',
'detector_type': 'offline',
'data_type': None,
'version': __version__
}
# fit OutlierAEGMM, infer threshold and compute scores
aegmm.fit(X, w_energy=w_energy, epochs=5, batch_size=1000, verbose=False)
aegmm.infer_threshold(X, threshold_perc=threshold_perc)
energy = aegmm.score(X)
perc_score = 100 * (energy <
aegmm.threshold).astype(int).sum() / energy.shape[0]
assert threshold_perc + 5 > perc_score > threshold_perc - 5
# make and check predictions
od_preds = aegmm.predict(X, return_instance_score=return_instance_score)
assert od_preds['meta'] == aegmm.meta
assert od_preds['data']['is_outlier'].shape == (X.shape[0], )
if return_instance_score:
assert od_preds['data']['is_outlier'].sum() == (
od_preds['data']['instance_score'] >
aegmm.threshold).astype(int).sum()
else:
assert od_preds['data']['instance_score'] is None
def init_od_aegmm(state_dict: Dict, aegmm: tf.keras.Model) -> OutlierAEGMM:
"""
Initialize OutlierAEGMM.
Parameters
----------
state_dict
Dictionary containing the parameter values.
aegmm
Loaded AEGMM.
Returns
-------
Initialized OutlierAEGMM instance.
"""
od = OutlierAEGMM(threshold=state_dict['threshold'], aegmm=aegmm)
od.phi = state_dict['phi']
od.mu = state_dict['mu']
od.cov = state_dict['cov']
od.L = state_dict['L']
od.log_det_cov = state_dict['log_det_cov']
if not all(
tf.is_tensor(_)
for _ in [od.phi, od.mu, od.cov, od.L, od.log_det_cov]):
logger.warning('Loaded AEGMM detector has not been fit.')
return od
threshold_net = tf.keras.Sequential([
InputLayer(input_shape=(seq_len, latent_dim)),
Dense(5, activation=tf.nn.relu)
])
# define model
inputs = tf.keras.Input(shape=(input_dim, ))
outputs = tf.keras.layers.Dense(2, activation=tf.nn.softmax)(inputs)
model = tf.keras.Model(inputs=inputs, outputs=outputs)
detector = [
AdversarialAE(threshold=threshold, model=model, **kwargs),
IForest(threshold=threshold),
Mahalanobis(threshold=threshold),
OutlierAEGMM(threshold=threshold,
gmm_density_net=gmm_density_net,
n_gmm=n_gmm,
**kwargs),
OutlierVAE(threshold=threshold,
latent_dim=latent_dim,
samples=samples,
**kwargs),
OutlierAE(threshold=threshold, **kwargs),
OutlierVAEGMM(threshold=threshold,
gmm_density_net=gmm_density_net,
n_gmm=n_gmm,
latent_dim=latent_dim,
samples=samples,
**kwargs),
OutlierProphet(threshold=.7, growth='logistic'),
SpectralResidual(threshold=threshold, window_amp=10, window_local=10),
OutlierSeq2Seq(input_dim,