Seminar on Machine Learning Systems
Overview
The seminar covers core concepts and ideas in the general area of machine learning systems, ranging from distributed and federated learning systems, DevOps systems for ML, life cycle and data management systems for MLs, etc. The focus will be to cover fundamental ideas on ML systems, with an emphasis on software systems and platforms.
The seminar will consist of student presentations based on a list of papers that will be provided at the beginning of the course. Presentations will be given in groups of four students. The four members of the group will each take the following roles:(1) Authors (2) Peer reviewers (3) Historians (4) Graduate students. Each group needs to choose 3 papers, and each member should choose different roles for each different papers.
The presentation is split into four parts (1) The ‘Historians’ start the discussion by setting the context, explaining the state of the art, the relevance, and timeliness of the larger research, and showing how and why the ‘Authors’ came up with their idea in the first place. (2) The ‘Authors’ then present and defend the paper’s methodology, results, and conclusions. (3) Next, the ‘Peer reviewers’ present their critique of the paper. (4) Finally, the ‘Graduate students’ explain how this paper could serve as inspiration for future research, and how they would take things a step further. Grades will be assigned based on the quality of the presentation, coverage of the topic including material not in the original papers, participation during the seminar, and ability to understand, present, and criticize the underlying technology.
Each team is requested to hand in a detailed review form before the day of the presentation, summarizing the paper, the critical analysis, and the questions prepared for the discussion.
Seminar Hours
Wednesday 14:15 - 16:00.
Room: IFW A 34