Going To Conferences

You should plan on going to at least one major conference per year. It is a lot of fun, especially after you get to know some of the people, and it is good for your career. You should plan your work ahead so that you have a paper ready when the deadline comes around. In the beginning it might make sense to go to a primary conference such as NIPS or AAAI or GECCO even if you don't have a paper there just to get an idea of what the atmosphere is like.

Which conference?

You should submit primarily to the best conferences in your area. The general machine learning conferences, such as NeurIPS, AAAI, IJCAI, ICLR, ICML, may have acceptance rates below 15% (see a list of recent acceptance rates). As a result, the papers in these conferences tend not to be very diverse, and it may take several rounds to get a paper accepted. Unless you have a result that is very mainstream machine learning, it may be better to send to a more specialized conference with a more reasonable acceptance rate. For instance, GECCO and Cognitive Science conferences have acceptance rates in the 30s---a publication there is still prestigious and likely to be of good quality, but the rate is high enough to include a diversity of papers.

IEEE conferences such as ICJNN, CEC, and WCCI, as well as Alife, Evostar, PPSN, ICDL etc. are possible alternatives: their acceptance rates tend to be a little too high (over 50%), which means they are not as prestigious and not all papers are that great. But they may be a good way to start, especially with a smaller initial result.

Sometimes small workshops (such as those at NeurIPS, GECCO, AAAI, and IJCAI, and the AAAI Spring and Fall Symposia, and IEEE Symposium Series on Computational Intelligence), with a focused topic are very good too, because it is easy to talk to the people in your area. Summer schools are also a very good way to meet both established researchers and other students. Papers at such venues aren't very prestigious, but prestige is not the point. Workshops are intended to showcase work in progress, and people go there to interact.

What to do at a conference?

The main activity at the conference is not to attend the talks, but to meet people. That's how you find out what's going on, get new ideas, and make the contacts that help you in your career. Sure, you go to those talks that seem the most interesting, but there's really only a limited number of those that you can absorb. So spend the rest of the time talking to people. The breaks, lunches, and receptions are the best times to do that, and most conferences have poster sessions that are a lot of fun (and often run past midnight...).

It is a good idea to be prepared before you go, i.e. look at the schedule and identify interesting talks, note who has which papers, and think of topics you might bring up with people when you meet them.

Preparing a conference talk

Conference presentations in AI are usually about 15 or 20 mins + 5 for questions. The talk should follow your paper for the most part, although it is perfectly fine to include new results obtained after the paper was accepted. You cannot present all the details in a short talk, and most people won't be able to follow a very detailed talk anyway. Try to get a few central ideas through as clearly as possible. See the page on making presentations for tips, tools, and guidelines. A few days before the conference we can schedule an NN-meeting where you can give a practice talk. We'll give you lots of feedback, and you can then refine the talk before you leave for the conference.

Preparing a poster

If you have a poster presentation, you should prepare it well ahead of time, and spend some effort into making it good. It is not sufficient to just paste your paper on the board! The department has a poster printer that allows you to print the whole poster on one piece of paper. You can then take it with you to the conference in a tube (a bit of a pain to take on the plane---sometimes you may be able to print it locally where the conference is).

The poster should have a large title across the top, with the author names and affiliations below, and the sections should be clearly organized e.g. on differently-colored background. Each section should have a large title, perhaps a figure, and one central point in a large font, and perhaps more details in a smaller font. You need the organization and titles because a person who walks to the poster should be able to immediately see what it is all about, you need the figures so that you can explain it to others, and you need the details for people who want to study your poster in detail when you are not there. Again, color will help organize and emphasize ideas on your poster a lot, so try to make use of it. They can be prepared in LaTex (there's an example in cs:/u/nn/tex/example/poster), Overleaf, Keynote, and other software. Take a look at a few previous posters by the NN Group (click on the "Slides(PDF)" link e.g. on SFN19 for LaTex, CogSci22 for Keynote (with source under "Other"))---they should make the description above concrete.

Presenting a poster is a little different from presenting a paper (i.e. giving a talk). People wonder by and make ask questions, and the discussion can take different directions. However, you need to prepare a "tour" of the poster as well, i.e. a presentation of a few minutes (5-10) that goes over the material in the poster in your own words. When a potential customer stops by, you can then ask them whether they would like to get the tour.

Travel Arrangements and Expenses

In order to get travel expenses paid (from a grant or otherwise), you need to turn in request for travel authorization well in advance. Usually you make your own arrangements (perhaps using the Concur travel management system) and get reimbursed after the trip. So make sure to keep the receipts!

Travel funding is usually limited so we have to be economical. Book nonrefundable airfares and register to the conference early so that you get a cheaper rate. It is usually a good idea to stay at the conference hotel---it makes it more convenient to participate---but you may need to share a room with another student to reduce the expense. Some conferences and the department also has partial travel grants that you should take advantage of.

Last modified: Sun Sep 25 17:27:48 PDT 2022