def test_algorithm_initializations():
def detectors(seed):
dets = [
AutoEncoder(num_epochs=1, seed=seed),
DAGMM(num_epochs=1, seed=seed),
DAGMM(num_epochs=1,
autoencoder_type=DAGMM.AutoEncoder.LSTM,
seed=seed),
LSTMAD(num_epochs=1, seed=seed),
LSTMED(num_epochs=1, seed=seed),
RecurrentEBM(num_epochs=1, seed=seed)
]
return sorted(dets, key=lambda x: x.framework)
RUNS = 1
seeds = np.random.randint(np.iinfo(np.uint32).max,
size=RUNS,
dtype=np.uint32)
output_dir = 'reports/experiments'
evaluators = []
outlier_height_steps = 1
for outlier_type in ['extreme_1', 'shift_1', 'variance_1', 'trend_1']:
announce_experiment('Outlier Height')
ev_extr = run_extremes_experiment(detectors,
seeds,
RUNS,
outlier_type,
steps=outlier_height_steps,
output_dir=os.path.join(
output_dir, outlier_type,
'intensity'))
evaluators.append(ev_extr)
ev_extr.plot_single_heatmap()
def run_experiments():
# Set the seed manually for reproducibility.
seeds = np.random.randint(np.iinfo(np.uint32).max,
size=RUNS,
dtype=np.uint32)
output_dir = 'reports/experiments'
evaluators = []
announce_experiment('Multivariate Datasets')
ev_mv = run_multivariate_experiment(detectors,
seeds,
RUNS,
output_dir=os.path.join(
output_dir, 'multivariate'))
evaluators.append(ev_mv)
for ev in evaluators:
ev.plot_single_heatmap()
def run_experiments():
# Set the seed manually for reproducibility.
seeds = np.random.randint(np.iinfo(np.uint32).max,
size=RUNS,
dtype=np.uint32)
output_dir = 'reports/experiments'
evaluators = []
outlier_height_steps = 10
for outlier_type in ['extreme_1', 'shift_1', 'variance_1', 'trend_1']:
announce_experiment('Outlier Height')
ev_extr = run_extremes_experiment(detectors,
seeds,
RUNS,
outlier_type,
steps=outlier_height_steps,
output_dir=os.path.join(
output_dir, outlier_type,
'intensity'))
evaluators.append(ev_extr)
announce_experiment('Multivariate Datasets')
ev_mv = run_multivariate_experiment(detectors,
seeds,
RUNS,
output_dir=os.path.join(
output_dir, 'multivariate'))
evaluators.append(ev_mv)
for mv_anomaly in [
'doubled', 'inversed', 'shrinked', 'delayed', 'xor',
'delayed_missing'
]:
announce_experiment(f'Multivariate Polluted {mv_anomaly} Datasets')
ev_mv = run_multivariate_polluted_experiment(detectors,
seeds,
RUNS,
mv_anomaly,
output_dir=os.path.join(
output_dir,
'mv_polluted'))
evaluators.append(ev_mv)
announce_experiment(
f'High-dimensional multivariate {mv_anomaly} outliers')
ev_mv_dim = run_multi_dim_multivariate_experiment(
detectors,
seeds,
RUNS,
mv_anomaly,
steps=20,
output_dir=os.path.join(output_dir, 'multi_dim_mv'))
evaluators.append(ev_mv_dim)
announce_experiment('Long-Term Experiments')
ev_different_windows = run_different_window_sizes_evaluator(
different_window_detectors, seeds, RUNS)
evaluators.append(ev_different_windows)
for ev in evaluators:
ev.plot_single_heatmap()