Federated Learning of Cohorts — Google’s cookie killer

Interest-based advertising without access to personal data

From this article, you’ll learn how Google is supposed to replace cookie-based tracking by using some federated learning concepts. You’ll also learn the details of algorithms that could shape the future of the online industry. I will show you how to reproduce this approach to segment users by their book interests.

Unless stated otherwise, all pictures in the article are by the author.

The success of the machine learning projects depends on access to high-quality and large-scale datasets. So far, the common methodology for creating models has been to collect user’s data in the central place and use it however you want.

In today’s web reality, it means relying on a variety of user tracking techniques, such as third-party cookies (and alternative storages) and device fingerprinting. But in a world of data leaks, GDPR, CCPA and increased data protection legislation inspired by those, this approach becomes very inconvenient. Safari and Firefox have built-in solutions for reducing cross-site tracking already. Nearly a year and a half ago, Chrome’s team announced that they are going to drop third-party cookies from their browser within 2 years.

🔔 Want more articles like this? Sign up here.