Steganografiya quyidagi sohalarda qo'llaniladi, lekin ular bilan cheklanmaydi

Quantification parameter q for regression

"""
number_labels = len(np.unique([floor(k) for k in y_train]))
# CLASSIFICATION MODELS ###

# Random Forest Classifier
if isinstance(self.model, RandomForestClassifier):
self.model.fit(X_train, y_train)
predictions = self.model.predict(ownership['inputs'])
if verify(
ownership['labels'],
predictions,
number_labels=number_labels,
metric=self.metric)['is_stolen']:
self.watermarked = True
return None

# Logistic Regression
if isinstance(self.model, LogisticRegression):
self.model.fit(X_train, y_train)
predictions = self.model.predict(ownership['inputs'])
if verify(
ownership['labels'],
predictions,
number_labels=number_labels,
metric=self.metric)['is_stolen']:
self.watermarked = True
return None

# SVC
elif isinstance(self.model, svm.SVC):
self.model.fit(X_train, y_train)
predictions = self.model.predict(ownership['inputs'])
if verify(ownership['labels'],
predictions,
number_labels=number_labels,
metric=self.metric)['is_stolen']:
self.watermarked = True

return None

# Ridge
elif isinstance(self.model, RidgeClassifier):
self.model.fit(X_train, y_train)
predictions = self.model.predict(ownership['inputs'])
if verify(ownership['labels'],
predictions,
number_labels=number_labels,
metric=self.metric)['is_stolen']:
self.watermarked = True

return None

# REGRESSION MODELS

# Random Forest Regressor