secure_learning.metrics module

Provides classification and regression metrics.

secure_learning.metrics.accuracy_score(y_real, y_pred)

Computes the accuracy of the predicted labels. Accuracy is computed as the ratio of all correctly predicted labels over the number of predicted labels.

Parameters:

• y_real (List[SecureFixedPoint]) – Real labels (-/+ 1)

• y_pred (List[SecureFixedPoint]) – Predicted labels (-/+ 1)

Return type:

SecureFixedPoint

Returns:

Accuracy of given predictions

secure_learning.metrics.adj_r2_score(y_real, y_pred, n_features)

Adjusted R-squared value for given predicted and real target values.

Parameters:

• y_real (List[SecureFixedPoint]) – Real target values

• y_pred (List[SecureFixedPoint]) – Predicted target values

• n_features (int) – Number of features

Return type:

SecureFixedPoint

Returns:

Adjusted R-squared value

secure_learning.metrics.f1_score(y_real, y_pred, pos_label)

F1-score for given predicted and real target labels.

Parameters:

• y_real (List[SecureFixedPoint]) – Real labels (-/+ 1)

• y_pred (List[SecureFixedPoint]) – Predicted labels (-/+ 1)

• pos_label (int) – Label to compute f1 score of

Raises:

SecureLearnValueError – pos_label value must be either -1 or 1

Return type:

SecureFixedPoint

Returns:

F1 score

secure_learning.metrics.mean_squared_error(y_real, y_pred)

Compute residual mean of squares. Residual mean of squares equals the mean of squares of deviations between predicted and real values.

Parameters:

• y_real (List[SecureFixedPoint]) – Real labels (-/+ 1)

• y_pred (List[SecureFixedPoint]) – Predicted labels (-/+ 1)

Return type:

SecureFixedPoint

Returns:

Residual mean of squares

secure_learning.metrics.mean_squared_model(y_real)

Compute explained mean of squares. Explained sum of squares equals the mean of squares of deviations from the mean.

Parameters:

y_real (List[SecureFixedPoint]) – Input

Return type:

SecureFixedPoint

Returns:

Explained mean of squares

secure_learning.metrics.precision_score(y_real, y_pred, pos_label)

Computes the precision of the predicted labels of category pos_label. Precision is computed as the ratio of all correctly predicted pos_label over the number of predicted pos_label. This is an indication how precise the predictions of pos_label are: given prediction pos_label, how likely is it that the true label is pos_label.

Parameters:

• y_real (List[SecureFixedPoint]) – Real labels (-/+ 1)

• y_pred (List[SecureFixedPoint]) – Predicted labels (-/+ 1)

• pos_label (int) – Label (value) to compute precision of

Raises:

SecureLearnValueError – pos_label value must be either -1 or 1

Return type:

SecureFixedPoint

Returns:

Precision of given predictions

secure_learning.metrics.r2_score(y_real, y_pred)

R-squared value for given predicted and real target values.

Parameters:

• y_real (List[SecureFixedPoint]) – Real target values

• y_pred (List[SecureFixedPoint]) – Predicted target values

Return type:

SecureFixedPoint

Returns:

R-squared value

secure_learning.metrics.recall_score(y_real, y_pred, pos_label)

Computes the recall of the predicted labels of category pos_label. Recall is computed as the ratio of all correctly predicted pos_label over the number of real pos_label. This is an indication how many actual pos_label we misclassified.

Parameters:

• y_real (List[SecureFixedPoint]) – Real labels (-/+ 1)

• y_pred (List[SecureFixedPoint]) – Predicted labels (-/+ 1)

• pos_label (int) – Label (value) to compute recall of

Raises:

SecureLearnValueError – pos_label value must be either -1 or 1

Return type:

SecureFixedPoint

Returns:

Recall of given predictions