+ Preventing ML models from exposing potentially sensitive information is a critical part of + using AI responsibly. To that end, differentially private stochastic gradient descent + (DP-SGD) is a modification to the standard stochastic gradient descent (SGD) algorithm + in machine learning.
+Models trained with DP-SGD have provable differential privacy (DP) + guarantees, mitigating the risk of exposing sensitive training data. Intuitively, a model + trained with differential privacy should not be affected by any single training example in + its data set. DP-SGD techniques can also be used in federated learning to provide user-level + differential privacy. You can learn more about differentially private deep learning in the original paper. +
- code_block: | + - code_block: |
import tensorflow as tf
- import PROJECT_NAME
+ from tensorflow_privacy.privacy.optimizers import dp_optimizer_keras
- # This is a short code snippet that shows off your project.
- # Launch a Colab notebook to run this example.
- print("Hello, PROJECT_NAME")
-
- {% dynamic if request.tld != 'cn' %}
- Run in a Notebook
- {% dynamic endif %}
+ # Select your differentially private optimizer
+ optimizer = tensorflow_privacy.DPKerasSGDOptimizer(
+ l2_norm_clip=l2_norm_clip,
+ noise_multiplier=noise_multiplier,
+ num_microbatches=num_microbatches,
+ learning_rate=learning_rate)
+
+ # Select your loss function
+ loss = tf.keras.losses.CategoricalCrossentropy(
+ from_logits=True, reduction=tf.losses.Reduction.NONE)
+
+ # Compile your model
+ model.compile(optimizer=optimizer, loss=loss, metrics=['accuracy'])
+
+ # Fit your model
+ model.fit(train_data, train_labels,
+ epochs=epochs,
+ validation_data=(test_data, test_labels),
+ batch_size=batch_size)
+
+
+ - classname: devsite-landing-row-100
+ - heading: TensorFlow Privacy
+ options:
+ - description-100
+ items:
+ - classname: devsite-landing-row-100
+ description: >
+ Tensorflow Privacy (TF Privacy) is an open source library developed by teams in Google + Research. The library includes implementations of commonly used TensorFlow Optimizers for + training ML models with DP. The goal is to enable ML practitioners using standard Tensorflow + APIs to train privacy-preserving models by changing only a few lines of code.
+The differentially private Optimizers can be used in conjunction with high-level APIs + that use the Optimizer class, especially Keras. Additionally, you can find differentially + private implementations of some Keras models. All of the Optimizers and models can be found + in the API Documentation.
- classname: devsite-landing-row-cards items: - - heading: "Introducing PROJECT_NAME" + - heading: "Introducing TensorFlow Privacy" image_path: /resources/images/tf-logo-card-16x9.png - path: https://blog.tensorflow.org + path: https://blog.tensorflow.org/2019/03/introducing-tensorflow-privacy-learning.html buttons: - label: "Read on TensorFlow blog" - path: https://blog.tensorflow.org - - heading: "PROJECT_NAME video" - youtube_id: 3d34Hkf7KXA + path: https://blog.tensorflow.org/2019/03/introducing-tensorflow-privacy-learning.html + - heading: "TensorFlow Privacy at TF Dev Summit 2020" + youtube_id: UEECKh6PLhI buttons: - label: Watch the video - path: https://www.youtube.com/watch?v=3d34Hkf7KXA - - heading: "PROJECT_NAME on GitHub" + path: https://www.youtube.com/watch?v=UEECKh6PLhI + - heading: "TensorFlow Privacy on GitHub" image_path: /resources/images/github-card-16x9.png - path: https://github.com/tensorflow/PROJECT_NAME + path: https://github.com/tensorflow/privacy buttons: - label: "View on GitHub" - path: https://github.com/tensorflow/PROJECT_NAME + path: https://github.com/tensorflow/privacy