Open Source

ML Applications

Facebook AI, OpenMined partner on new PyTorch privacy and machine learning courses

November 12, 2020

Facebook AI is partnering with OpenMined, an open source community focused on privacy in artificial intelligence and machine learning (ML), to offer developers a series of educational courses called The Private AI Series, based on PyTorch.

ML models, especially those that leverage sensitive data, have a responsibility to preserve data privacy. This means ensuring not only that sensitive data is protected from malicious hackers, but also that people’s information, the information the models deliver, and the models themselves cannot be accessed by unauthorized parties.

This new PPML series is designed for all levels, from beginners to advanced-level developers, to provide an overview of privacy-related technologies and to teach the skills to implement and analyze them. This series has two goals: to provide the broadest possible awareness and education around both PPML concepts and hands-on development, and to promote key ML and PPML technologies. Students will use PyTorch in their projects as well as PySyft, a Python library to develop secure and private ML models.

The series will be broken into four courses:

Awareness: Participants will learn the ins and outs of privacy-enhancing technology (PET), why privacy matters, and how PET is changing the business landscape. This course will give students an understanding of key privacy-related technologies and terms, and will empower them to understand how their own industries will be transformed by PETs, the opportunities that will be created, and how PETs will disrupt business models.

Foundation: This course will go under the hood and teach students how to design and evaluate high-level systems that use privacy PETs. After this course, students will be able to comprehend technical PET papers, write implementations based on them, and design systems that leverage PETs. Topics covered will include federated learning, split learning, differential privacy, homomorphic encryption, cryptographic signatures, public key technology, and more.

Cross-Enterprise Statistics and Federated Learning: This practitioner course will give students an understanding of how to use PyTorch and PySyft to perform cross-organizational analysis. Among the topics covered, students will learn how to search for data that cannot be seen; clean and extract, transform, and load (ETL) data without being able to analyze it; train models and perform statistical analysis using remote execution tools; and work with models and data across multiple organizations simultaneously.

Federated Statistics and Learning on Web and Mobile: This course will cover how to build web and mobile apps that store data on devices running Android, iOS, and React.js. Students will use PyTorch to perform distributed database queries and train models using federated learning. By the end of the course, participants will understand how to build an application with a distributed, on-device database; build an application while training a federated learning model on a device; and leverage private-set intersection for private queries between a server and a client.

The Awareness course will begin in January 2021. Anyone interested in registering can visit for more information.