What Is Federated Learning?

Chapter 3. Case Studies

Now that you’re familiar with federated learning, we will examine some real-life case studies of this technology being used today. In this chapter, we’ll look at examples of both cross-device and cross-silo federated learning.

Cross-Device Case Studies

Let’s look at three case studies that illustrate some interesting uses for federated learning and analytics in the real world.

Next Word Prediction in a Mobile Keyboard

Gboard, the Google keyboard, is a virtual keyboard for touchscreen mobile devices with support for more than 600 language varieties and over 1 billion installs as of 2019 (Hard et al., 2019). Gboard employs a variety of methods to improve user experience and keyboard features, including federated learning with differential privacy.

To see how Gboard can use differential privacy and federated learning, let’s walk through a next word prediction example. Next word prediction aims to improve the user experience of text entry by predicting the next word a user will type given the most relevant phrase or sequence of words. For example, using a next word prediction model, Gboard can predict that if the user has just typed the phrase “I love you,” they are likely to continue by entering “so,” “too,” or a kissing or heart emoji. Thus, it provides these likely options as suggestions a user can easily add to their text while typing, as shown in Figure 3-1.

Figure 3-1. Gboard’s predictions for likely words or phrases

When Gboard shows a user these ...