Skip to main content Link Menu Expand (external link) Document Search Copy Copied
Federated Learning

Overview

As data from various sources and stakeholders becomes central to powering machine learning pipelines, the need for collaborative learning techniques is more crucial than ever. Federated learning, for example, is poised to revolutionize the field by facilitating secure and coordinated learning across different parties and heterogeneous data sets. To achieve this, innovative strategies must be developed to enhance the accuracy and efficiency of learning with isolated data; reduce risks while safeguarding data privacy and ownership; and integrate social and economic incentives that encourage data sharing and establish reliable cooperative learning frameworks. The objective of this course is to introduce the state-of-the-art systems, algorithms, and applications in federated learning. Topics to be covered include but are not limited to:

  • Distributed optimization
  • Data heterogeneity in federated learning
  • System heterogeneity in federated learning
  • Security and privacy in federated learning
  • Federated learning with emerging foundation models

Grading Policy (subject to change)

The course requirements include active participation in and leadership of discussion sessions, as well as the completion of a course project. The grading breakdown is as follows:

  • Paper Review (30%): Each week, non-presenting students are required to review one of the two papers under discussion. The review should follow the NeurIPS conference format, with specific instructions and a template provided in Canvas. Further details will be explained during the lecture.

  • Paper Presentation and Discussion (30%): Every student will present two selected papers over the course of the semester and lead the discussion for each. Non-presenting students must prepare at least one question for each paper discussed.

  • Team Project (40%): Teams can consist of up to 3 students, with the project grade divided as follows:

    • Proposal (10%): A two-page project proposal is due by October 13th. A template will be available on Canvas.
    • Proposal Presentation (10%): Each team will present their project during the week of October 14th.
    • Final Project Report (10%): A nine-page final report, adhering to the NeurIPS submission template, must be submitted by December 9th.
    • Final Project Presentation (10%): Teams will present their final projects on December 4th and December 9th.

Lecture Information

  • This course meets on Mon/Wed 3:30 PM-4:45 PM in EGR 0135.
  • The prerequisites for this course are: ENEE436 or CMSC422; or students who have taken courses with comparable content may contact the Department.

Textbook (optional)

  • There are no required textbooks for this course; all required readings will be in the form of papers.

Office Hours

  • Instructor office hours: Monday 12:30-1:00 PM and by appointments
  • TA office hours: TBD

All students are presumed to be aware of the UMD policy on academic integrity.