Implementation of Differentially Private Logistic Regression.

PiperOrigin-RevId: 458266079

tensorflow_privacy/privacy/logistic_regression/datasets.py

@@ -29,7 +29,7 @@ Includes two types of datasets:
 """
 
 import dataclasses
-from typing import Tuple
+from typing import Tuple, Optional
 
 import numpy as np
 from sklearn import preprocessing
@@ -46,9 +46,13 @@ class RegressionDataset:
     labels: array of shape (num_examples,) containing the corresponding labels,
       each belonging to the set {0,1,...,num_classes-1}, where num_classes is
       the number of classes.
+    weights: dimension by num_classes matrix containing coefficients of linear
+      separator, where dimension is the dimension and num_classes is the number
+      of classes.
   """
   points: np.ndarray
   labels: np.ndarray
+  weights: Optional[np.ndarray]
 
 
 def linearly_separable_labeled_examples(
@@ -71,7 +75,7 @@ def linearly_separable_labeled_examples(
   points = preprocessing.normalize(points_non_normalized)
   # Compute labels.
   labels = np.argmax(np.matmul(points, weights), axis=1)
-  return RegressionDataset(points, labels)
+  return RegressionDataset(points, labels, weights)
 
 
 def synthetic_linearly_separable_data(
@@ -122,5 +126,5 @@ def mnist_dataset() -> Tuple[RegressionDataset, RegressionDataset]:
       (num_test, -1))
   train_points = preprocessing.normalize(train_points_non_normalized)
   test_points = preprocessing.normalize(test_points_non_normalized)
-  return (RegressionDataset(train_points, train_labels),
-          RegressionDataset(test_points, test_labels))
+  return (RegressionDataset(train_points, train_labels, None),
+          RegressionDataset(test_points, test_labels, None))

tensorflow_privacy/privacy/logistic_regression/multinomial_logistic_test.py

@@ -33,7 +33,7 @@ class MultinomialLogisticRegressionTest(parameterized.TestCase):
                                            tolerance):
     (train_dataset, test_dataset) = datasets.synthetic_linearly_separable_data(
         num_train, num_test, dimension, num_classes)
-    accuracy = multinomial_logistic.logistic_objective_perturbation(
+    _, accuracy = multinomial_logistic.logistic_objective_perturbation(
         train_dataset, test_dataset, epsilon, delta, epochs, num_classes, 1)
     # Since the synthetic data is linearly separable, we expect the test
     # accuracy to come arbitrarily close to 1 as the number of training examples
@@ -67,11 +67,9 @@ class MultinomialLogisticRegressionTest(parameterized.TestCase):
                           num_microbatches, clipping_norm):
     (train_dataset, test_dataset) = datasets.synthetic_linearly_separable_data(
         num_train, num_test, dimension, num_classes)
-    accuracy = multinomial_logistic.logistic_dpsgd(train_dataset, test_dataset,
-                                                   epsilon, delta, epochs,
-                                                   num_classes, batch_size,
-                                                   num_microbatches,
-                                                   clipping_norm)
+    _, accuracy = multinomial_logistic.logistic_dpsgd(
+        train_dataset, test_dataset, epsilon, delta, epochs, num_classes,
+        batch_size, num_microbatches, clipping_norm)
     # Since the synthetic data is linearly separable, we expect the test
     # accuracy to come arbitrarily close to 1 as the number of training examples
     # grows.

tensorflow_privacy/privacy/logistic_regression/single_layer_softmax.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 """Implementation of a single-layer softmax classifier."""
 
-from typing import List, Optional, Union
+from typing import List, Optional, Union, Tuple, Any
 
 import tensorflow as tf
 from tensorflow_privacy.privacy.logistic_regression import datasets
@@ -28,7 +28,7 @@ def single_layer_softmax_classifier(
     loss: Union[tf.keras.losses.Loss, str] = 'categorical_crossentropy',
     batch_size: int = 32,
     kernel_regularizer: Optional[tf.keras.regularizers.Regularizer] = None
-) -> List[float]:
+) -> Tuple[Any, List[float]]:
   """Trains a single layer neural network classifier with softmax activation.
 
   Args:
@@ -63,4 +63,5 @@ def single_layer_softmax_classifier(
       epochs=epochs,
       validation_data=(test_dataset.points, one_hot_test_labels),
       verbose=0)
-  return history.history['val_accuracy']
+  weights = model.layers[0].weights
+  return weights, history.history['val_accuracy']

tensorflow_privacy/privacy/logistic_regression/single_layer_softmax_test.py

@@ -31,10 +31,9 @@ class SingleLayerSoftmaxTest(parameterized.TestCase):
                                 num_classes, tolerance):
     (train_dataset, test_dataset) = datasets.synthetic_linearly_separable_data(
         num_train, num_test, dimension, num_classes)
-    accuracy = single_layer_softmax.single_layer_softmax_classifier(
+    _, accuracy = single_layer_softmax.single_layer_softmax_classifier(
         train_dataset, test_dataset, epochs, num_classes, 'sgd')
     self.assertAlmostEqual(accuracy[-1], 1, delta=tolerance)
 
-
 if __name__ == '__main__':
   unittest.main()