Instructor: Pranay Sharma
TA: TBD
Time: Tuesdays and Fridays, 2.00-3.30 pm
Room: LT 006
Office Hours: Fridays 3.30-4.30 pm
Course Description
Modern machine learning applications often rely on data generated and/or collected at the edge-devices (phones, smart watches, sensors, etc.). However, bringing this data to a central location (often called a server) is often infeasible (high communication cost) or undesirable (privacy concerns). Distributed learning is essential in such applications to efficiently learn large-scale ML models.
This course will begin with some basics of optimization. This will be followed by an extensive discussion on federated learning – a general framework under the worker-server network architecture. The last few lectures will focus on distributed learning with agents connected in a more general network topology.