3.0

credits

Average Course Rating

Federated Learning (FL) is an area of machine learning where data is distributed across multiple devices and training is performed without exchanging the data between devices. FL can be contrasted with classical machine learning settings when data is available in a central location. As such, FL faces additional challenges and limitations such as privacy and communication. For example, FL may deal with questions of learning from sensitive data on mobile devices while protecting privacy of individual users and dealing with low power and limited communication. As a result, FL requires knowledge of many interdisciplinary areas such as differential privacy, distributed optimization, sketching algorithms, compression and more. In this course students will learn basic concepts and algorithms for FL and federated analytics, and gain hands-on experience with new methods and techniques. Students will gain understanding in reasoning about possible trade-offs between privacy, accuracy and communication. [Analysis] Required:: 433/633 (Algo), 475/675 (ML) or 482/682 (ML: DL), linear algebra, probability