Topics in Optimization for Machine Learning

Seminar: Topics in Optimization for Machine Learning

Optimization lies at the heart of many machine learning algorithms. This seminar teaches an overview of modern mathematical optimization methods, for applications in machine learning and data science. In particular, we will discuss the theoretical basics of stochastic optimization, scalability of algorithms to large datasets, and challenges in distributed optimization, such as for instance in decentralized or federated machine learning. We will cover a set of foundational papers, but also a selection of recent publications. 

Previous attendance of the summer term lecture "Optimization for Machine Learning" is recommended, but not required.

Organization

In this seminar, students will learn to present, discuss, and summarize papers in different areas of optimization for machine learning. Specifically, each student will get a single topic assigned to them, consisting of two papers (a lead and follow-up paper). Each student will

• write a short seminar paper on the topic assigned to them, for which the two papers on the topic serve as the starting point;
• prepare a presentation on the topic assigned to them and present the paper to the other students;
• read the papers assigned to the other students and prepare questions to ask to the presenter of this paper/topic;
• there will be an additional writing assignment that can be chosen from the two tasks below at the beginning of the semester (please indicate your preference when applying for the seminar):
  • (A) write three short reviews on papers from a different topic. The reviews will be shared among the group (in particular with the presenter of the topic).
  • (B) write a research report on your topic. The goal is to identify an open research question in the assigned topic area and trying to work towards a (partial) solution of the identified question.

Important Dates (TBA)

• Kick-off meeting 7.11.22, 16.15h-17.00h (to be held online, via zoom).
• Voting for topics deadline: 11.11.22 (end of day)
• The presentations will take place between end of November - January (hybrid session format). Tentatively on Monday 4-6pm (we need to fix 6-7 dates, depending on number of participants). Participation in the presentations is mandatory.
• Mandatory feedback round/practice talk at least two weeks before the presentation (arrange exact time with supervisor).
• Hand-in of report: TBA.

Deliverables

• Questions: (10% of your final grade): submit a question for each presentation slot (except when you are presenting).

• Option A: 3 short reviews: (each contributes 10% of your final grade): Write a short review (max 2 page) on one of the papers (not the one that you are presenting) that addresses the following questions:

  1. What is the problem addressed by the paper?
  2. What was done before, and how does the paper improve on previous work?
  3. What are the strengths and the limitations of the techniques in the paper
  4. What part of the paper was difficult to understand?
  5. What are possible improvements or extensions of the techniques in the paper?

   

  Option B: a short scientific report on a research question: (30% of your final grade). (Format and length to be discussed with supervisor).

• Presentation: (30%). You will prepare and deliver a 30 min presentation (followed by 15 mins question/discussion) of the paper assigned to you. You will have the possibility to get feedback on your slides before the presentation.

• Seminar Paper: (30%) You will write a seminar paper on the topic that you have presented. It must not be longer than 6 pages, not counting references and appendices. Note that appendices are not meant to provide information that is absolutely necessary to understand the paper, but rather to provide auxiliary material. Papers can be shorter, but in general the provided page limit is a good indicator of how long a paper should be.