Hi all,

I've aggregated some tips and resources to help people choose a topic for a final project. I tried to focus on tips that would be useful in general when doing research and not only for the current project.

The best option might be to investigate some topic in which you already have an interest. Try to think of a topic that interests you, and how it might relate to learning theory. Skim through some papers related to this topic. Try to isolate a question that is specific enough to be addressed in this project, and still captures some important aspect of your topic.
To search for papers, I suggest using Google scholar. It has some nice features:
- If you find a paper that interests you, you can search for "Related articles". For example, here is a search result for papers related to Valiant's 1984 paper that introduced PAC learning. As you scroll through the first few pages of this search result, you'll see a lot of interesting ideas related to learning theory.
- You can see a list of all the papers that cite a particular paper. Once you find a paper that looks interesting, use this feature to see what further work has been done on the same topic.
- If you create a profile, you can even configure the site to display/send you emails with recommendations for papers on topics or by particular authors that interest you.
If you want to view a paper that is pay-walled, you can almost always access it for free using the UC Berkeley EZProxy. It's easy to set up in the browser, and has worked great for me.
Here are some venues that publish a lot of good papers on learning theory:
- The Annual Conference on Computational Learning Theory (COLT). Last two conferences: 2020, 2019.
- International Conference on Algorithmic Learning Theory (ALT). Last conference: 2020.
- International Conference on Machine Learning (ICML). All previous conferences: here (click on "papers", "tutorials", etc.).
Proceedings of Machine Learning Research. Contains all the papers published in ALT, COLT, ICML, NeurIPS, and more.
DBLP is a very good database of papers in computer science. In particular, you can use it to find lists of papers from a particular conference. For instance, here are all the papers ever published in COLT, going back all the way to 1988. To see all the papers for a particular year, click the view icon -> table of contents (e.g., all papers from COLT 1988).
I'd recommend going over the table of contents in BDSS and MRT books.
Skim through the bibliography in the course syllabus.
Check out the homepages and publications of some researchers in learning theory (this is a very partial and arbitrary list): Sanjeev Arora, Nina Balcan, Peter Bartlett, Shay Ben-David, Avrium Blum, ‪Nicolo Cesa-Bianchi‬, Amit Daniely‬, Constantinos Daskalakis, Ilias Diakonikolas, Moritz Hardt, ‪Elad Hazan‬, ‪Michael Kearns, Adam Klivans, ‪John Langford‬, Nick Littlestone, ‪David McAllester, Shay Moran, Ben Recht, ‪Robert Schapire‬, Ohad Shamir, Nati Srebro, Gregory Valiant, Leslie Valiant, Manfred Warmuth, Amir Yehudayoff.
Schedule an advisory.

Best,

Jonathan