def get_bu(model, X_test, X_test_noisy, X_test_adv):
"""
Get Bayesian uncertainty scores
:param model:
:param X_train:
:param Y_train:
:param X_test:
:param X_test_noisy:
:param X_test_adv:
:return: artifacts: positive and negative examples with bu values,
labels: adversarial (label: 1) and normal/noisy (label: 0) examples
"""
print('Getting Monte Carlo dropout variance predictions...')
uncerts_normal = get_mc_predictions(model, X_test,
batch_size=args.batch_size) \
.var(axis=0).mean(axis=1)
uncerts_noisy = get_mc_predictions(model, X_test_noisy,
batch_size=args.batch_size) \
.var(axis=0).mean(axis=1)
uncerts_adv = get_mc_predictions(model, X_test_adv,
batch_size=args.batch_size) \
.var(axis=0).mean(axis=1)
print("uncerts_normal:", uncerts_normal.shape)
print("uncerts_noisy:", uncerts_noisy.shape)
print("uncerts_adv:", uncerts_adv.shape)
## skip the normalization, you may want to try different normalizations later
## so at this step, just save the raw values
# uncerts_normal_z, uncerts_adv_z, uncerts_noisy_z = normalize(
# uncerts_normal,
# uncerts_adv,
# uncerts_noisy
# )
uncerts_pos = uncerts_adv
uncerts_neg = np.concatenate((uncerts_normal, uncerts_noisy))
artifacts, labels = merge_and_generate_labels(uncerts_pos, uncerts_neg)
return artifacts, labels
def get_bu(model, X_test, X_test_noisy, X_test_adv):
"""
Get Bayesian uncertainty scores
:param model:
:param X_train:
:param Y_train:
:param X_test:
:param X_test_noisy:
:param X_test_adv:
:return: artifacts: positive and negative examples with bu values,
labels: adversarial (label: 1) and normal/noisy (label: 0) examples
"""
print('Getting Monte Carlo dropout variance predictions...')
uncerts_normal = get_mc_predictions(model, X_test,
batch_size=args.batch_size) \
.var(axis=0).mean(axis=1)
uncerts_noisy = get_mc_predictions(model, X_test_noisy,
batch_size=args.batch_size) \
.var(axis=0).mean(axis=1)
uncerts_adv = get_mc_predictions(model, X_test_adv,
batch_size=args.batch_size) \
.var(axis=0).mean(axis=1)
print("uncerts_normal:", uncerts_normal.shape)
print("uncerts_noisy:", uncerts_noisy.shape)
print("uncerts_adv:", uncerts_adv.shape)
## skip the normalization, you may want to try different normalizations later
## so at this step, just save the raw values
# uncerts_normal_z, uncerts_adv_z, uncerts_noisy_z = normalize(
# uncerts_normal,
# uncerts_adv,
# uncerts_noisy
# )
uncerts_pos = uncerts_adv
uncerts_neg = np.concatenate((uncerts_normal, uncerts_noisy))
artifacts, labels = merge_and_generate_labels(uncerts_pos, uncerts_neg)
return artifacts, labels