6.853 Topics in Algorithmic Game Theory: AGT and Data Science
Spring 2021
Instructors: Konstantinos Daskalakis, Vasilis Syrgkanis
TAs: Maxwell K Fishelson, Noah Golowich
Lecture: WF1-2.30 (VIRTUAL)
Information:
This is a research-oriented course on game theory, data science and their interface. Unlike prior offerings of the class, this iteration will place a larger emphasis on research topics, comprising an introductory segment presenting the fundamentals, followed by student presentations of modern research topics. A main focus will be a semester-long project implemented under the guidance of the class staff.
The scope of the class is game theory and mechanism design, and the interaction of these topics with online learning, machine learning, econometrics and causal inference. The class is suitable for graduate students or advanced undergraduates with a background in Machine Learning and Algorithms. In particular, mathematical maturity and familiarity with algorithm analysis and learning theory will be assumed.
Class time will consist of lectures, student presentations, and group project meetings. Lectures will consist of instructor and student presentations and will happen live on Zoom. Lecture attendance is required and participation counts toward the grade. For students in distant time zones where lecture attendance is difficult, student presentations can be pre-recorded, and accompanied by a written review of the material presented.
Registration might be limited. Please register for the class and complete the following form:
https://forms.gle/ez5N7P7ayAoepCyD8
Announcements
Class Evaluation Reminder
Dear all:This is a reminder about filing your class evaluation report. Currently only 2 out of 15 eligible to respond have submitted one. I know everyone is busy with end of semester work but this is important to do.
Thanks,
Costis
--------
Dear all:
It has been great attending the project presentations this week, and seeing the great progress you all made, as well as the fruitful directions for research going forward. We are looking forward to the last three presentations on Wednesday, May 19th!
In the meantime, it would be great if you could all spend a few minutes to fill in the anonymous class evaluation form at the link below:
https://eduapps.mit.edu/subjeval/studenthome.htm
This year the format of the class (half lecturer + half student-lead presentations) and the emphasis of the learning outcomes (projects) was different than typical years. We would be especially interested about your thoughts regarding this format and possible improvements.
Best,
Costis (for the class staff)
Announced on 18 May 2021 6:17 p.m. by Konstantinos Daskalakis
Class Evaluation
Dear all:It has been great attending the project presentations this week, and seeing the great progress you all made, as well as the fruitful directions for research going forward. We are looking forward to the last three presentations on Wednesday, May 19th!
In the meantime, it would be great if you could all spend a few minutes to fill in the anonymous class evaluation form at the link below:
https://eduapps.mit.edu/subjeval/studenthome.htm
This year the format of the class (half lecturer + half student-lead presentations) and the emphasis of the learning outcomes (projects) was different than typical years. We would be especially interested about your thoughts regarding this format and possible improvements.
Best,
Costis (for the class staff)
Announced on 14 May 2021 4:00 p.m. by Konstantinos Daskalakis
Final Presentations and Report
Dear all:We are looking forward to the final presentations, starting tomorrow, according to the following schedule.
- Wednesday, May 12th:
Auctions with side information in high-dimensions and multiple
items
Angelos Pelecanos, Aristofanis Rontogiannis, Angelos Assos
Adversarial robustness in GMM problems
Yilun Zhou, Jun Wan, Hanshen Xiao
- Friday May, 14th:
Convergence of no-regret in first price auctions
Ali Haberman, Andy Haupt
Information gathering and matching markets
Boyu Liu, Austin Edelman
Adversarial approaches to balancing in treatment effect
estimation from observational data
Daniel Aronoff, Violet Felt
- Wednesday May, 19th:
Last-iterate convergence in over-parameterized GANs
Elbert Du
CLS Completeness
Julian Viera
Optimal Revenue from Few Samples
Md Sanzeed Anwar
Final presentations should take roughly 30 minutes (including
anticipated questions from the audience). Please respect this
timeframe as we have a packed schedule. The larger groups,
presenting tomorrow, have a bit more slack. When structuring your
final presentation, it is a good idea to make sure to include the
following components:
- A clear definition of the problem/model you are considering
- A description of the literature it fits in, and what is known
about it
- A discussion of the motivation & novelty of your question, as
well as the challenges in answering it
- A description of your results, proofs, simulations
- A description of the limitations of your approach, outstanding
challenges, future goals
For your final submission, you have two options:
- Submit the slides you used for your presentation, together with supplementary slides that give more details about your proofs/simulations/ideas/future plans
-Submit the slides you used for your presentation, together with a report that addresses the above points. The report should not be more than 8 pages in length (fullpage, 11pt font) including references, a title, and an abstract, and it may include an unbounded in length supplementary material section.
Finally, there is an automatic (by MIT) deadline of May 20th for the above, but if you find this timeframe too tight, please write to us and we will figure it out!
Best regards, and looking forward to the presentations!
Costis (for the staff)
Announced on 11 May 2021 3:10 p.m. by Konstantinos Daskalakis
CLASS CANCELLED FOR TODAY
Due to unforseeable circumstances, we are forced to cancel today’s presentation and class as a whole. We will continue with our planned schedule on Wednesday with final presentations.Announced on 07 May 2021 10:54 a.m. by Maxwell K Fishelson
Lecture this Friday *will* take place
Hi all,Contrary to what was announced in class today, there *will* be lecture this Friday 4/2 (at the normal time), to wrap up the topics discussed today on causality.
Announced on 31 March 2021 6:10 p.m. by Noah Golowich
