def init_od_s2s(state_dict: Dict,
seq2seq: tf.keras.Model) -> OutlierSeq2Seq:
"""
Initialize OutlierSeq2Seq.
Parameters
----------
state_dict
Dictionary containing the parameter values.
seq2seq
Loaded seq2seq model.
Returns
-------
Initialized OutlierSeq2Seq instance.
"""
seq_len, n_features = state_dict['shape'][1:]
od = OutlierSeq2Seq(n_features,
seq_len,
threshold=state_dict['threshold'],
seq2seq=seq2seq,
latent_dim=state_dict['latent_dim'],
output_activation=state_dict['output_activation'])
return od
def test_seq2seq(seq2seq_params):
# OutlierSeq2Seq parameters
n_features, seq_len, threshold, threshold_perc, return_instance_score, \
return_feature_score, outlier_perc, outlier_type = seq2seq_params
# create artificial sine time series
X = np.sin(np.linspace(-50, 50, 10000)).astype(np.float32).reshape((-1, n_features))
# create outliers for threshold and detection
X_threshold = inject_outlier_ts(X, perc_outlier=100 - threshold_perc, perc_window=10, n_std=10., min_std=9.).data
X_outlier = inject_outlier_ts(X, perc_outlier=100 - threshold_perc, perc_window=10, n_std=10., min_std=9.).data
# define architecture
od = OutlierSeq2Seq(n_features, seq_len, threshold=threshold, latent_dim=latent_dim)
if threshold is None:
assert od.threshold == 0.
else:
assert od.threshold == threshold
assert od.meta == {'name': 'OutlierSeq2Seq', 'detector_type': 'offline', 'data_type': 'time-series',
'version': __version__}
# fit OutlierSeq2Seq
od.fit(X, epochs=2, verbose=False)
# create some outliers and infer threshold
od.infer_threshold(X_threshold, threshold_perc=threshold_perc)
# compute scores and check ranges
fscore, iscore = od.score(X_threshold, outlier_perc=outlier_perc)
if isinstance(od.threshold, np.ndarray):
perc_score = 100 * (fscore < 0).astype(int).sum() / iscore.shape[0] / n_features
assert threshold_perc + 5 > perc_score > threshold_perc - 5
# create outliers and make predictions
od_preds = od.predict(X_outlier,
return_instance_score=return_instance_score,
return_feature_score=return_feature_score,
outlier_type=outlier_type,
outlier_perc=outlier_perc)
assert od_preds['meta'] == od.meta
if outlier_type == 'instance':
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'] > 0).astype(int).sum()
elif outlier_type == 'feature':
assert od_preds['data']['is_outlier'].shape == X.shape
if return_feature_score:
assert od_preds['data']['is_outlier'].sum() == (od_preds['data']['feature_score'] > 0).astype(int).sum()
if return_feature_score:
assert od_preds['data']['feature_score'].shape == X.shape
else:
assert od_preds['data']['feature_score'] is None
if return_instance_score:
assert od_preds['data']['instance_score'].shape == (X.shape[0],)
else:
assert od_preds['data']['instance_score'] is None
def seq2seq(data, df):
X_train = np.array(data).astype(np.float32)
# initialize outlier detector
n_features = 5
seq_len = 50
od = OutlierSeq2Seq(n_features, seq_len, threshold=None, latent_dim=20)
# train
od.fit(X_train[:7200], epochs=10, verbose=False)
od.infer_threshold(X_train[:7200], threshold_perc=95)
od_preds = od.predict(
X_train[:7200],
outlier_type='instance', # use 'feature' or 'instance' level
return_feature_score=True, # scores used to determine outliers
return_instance_score=True)
pred = (od_preds['data']['is_outlier'] == 1)
a = (df.iloc[:7200]).loc[pred]
return a
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,
seq_len,
threshold=threshold,
threshold_net=threshold_net,
latent_dim=latent_dim)
]
n_tests = len(detector)
@pytest.fixture
def select_detector(request):
return detector[request.param]
@pytest.mark.parametrize('select_detector',
list(range(n_tests)),
indirect=True)
def test_save_load(select_detector):
