Wednesday, September 10, 2008

SODA 2008 accepted papers list is out!

(Posted by request from Claire Mathieu.)

SODA'09 list of accepted papers is now available. The list also includes abstracts of the papers.
  1. Abstracts of the paper! This is great- I hope FOCS, STOC, COMPLEXITY, and OTHERS do that in the future. Makes it easier to tell if I want to download the paper ahead of time.
  2. Authors who got in CONGRADS!
  3. Authors who got in: make your conference version well written. Make sure that the casual reader knows what you did from the intro. Hilow complete do the proofs have to be? If you don't plan on writing a journal version (shame on you?) then make sure the proofs really are complete.
  4. If the conference version has complete proofs then is it worth getting out a journal version? In one sense you are supposed to since it gets refereed. Also, most Schools count journals more than conferences for promotion. But does refereeing really help? See this and other essay by Dr. Z for some interesting opinions.
  5. SODA has parallel Sessions. What are natural splits? Do people working on Approx algorithms not go to talks by people working on Randomized algorithms? I doubt that. Readers: what are your ideas on how to do parallel sessons? For SODA and in general.
  6. How many members of the COMPLEXITY community go to SODA? How many members of the SODA community to to COMPLEXITY? How can we define this question? We could see how many people who have been to 3 of the last 6 of conference X go to conference Y. I would guess that more COMPLEXITY people go to SODA then SODA people go to COMPLEXITY.


  1. ICML had abstracts, full papers AND discussion forum on the web, free of charge (and in html, not ascii). Happy to see that SODA is slowly, but certainly, advancing towards better practices.

  2. parallel sessions, length of the conference and the location of the conferences are always discussed at the end of each conference.
    Some conferences are held at the same location every year. Wondering - do people do get bored of the same location ?

  3. Parallel sessions are a failed model - too much work for the organizers, too much hassle for the attendees, who are constantly flipping through the program to see which talks they can miss/attend, shuttling between rooms, etc.

    A simple voting system that weighs votes by PC members and online registrants could be used to select roughly 30 (of the 243(!) accepts) for 30-min presentations - that's five hrs of talks for three days @ 30 mins per paper. The remaining 210 or so papers can be divided into three 2-hr poster sessions of 70 papers each - this would amply cater to the people long tail of papers that a small number of people are interested in. Throw in a few beers along with the posters, and you don't need the archaic "reception" event that several conferences still have - you can have a full party!

    Plus, limited seated time and more mingling time (on their feet) will keep the participants healthier :-)

  4. Parallel sessions are a failed model.

    This reminds me of the Yogi Berra aphorism on a trendy restaurant: "nobody eats there, it's too crowded".

    Most conferences within CS and otherwise run parallel sessions and they are thriving but somehow they are "a failed model" whereas stagnant single track conferences are great successes, I guess.

  5. That should be CONGRATS.

  6. That should be CONGRATS.

  7. Most conferences within CS and otherwise run parallel sessions and they are thriving

    Whether they are "thriving" is a matter of some debate.

    Consider the case of expensive commercial journals. Many people think the system is in terrible trouble: it's morally bankrupt, open access publishing is growing in popularity (and is fully equal in quality), and many library budgets won't even be able to pay for expensive journals much longer. Nevertheless, you could make a case that these journals are thriving. Profits are sky-high, expensive journals are proliferating wildly, most researchers cheerfully publish in them, and any collapse of the system is clearly years in the future. So in a sense expensive commercial journals are thriving, but I nevertheless consider them a failed model that will eventually be superceded.

    In any case, I don't entirely agree with the commenter who said parallel sessions are a failed model. However, saying parallel sessions are very popular today isn't a strong counterargument.

  8. 2. Authors who got in CONGRADS!
    3. Hilow complete do the proofs have to be?

    And then this?
    Authors who got in: make your conference version well written.

    How about we clean our own houses first?

  9. I prefer parallel sessions, but there should be better scheduling. Why not post abstracts/papers on the web and have registrants score talks they want to attend? Dump the anonymized ratings list on the web and have a week long contest to see who can schedule a minimum conflict assignment of talks. With sites like and two scripts this would be minimal effort on the PC's part.

  10. I added an html version of the list of accepted papers, compliment of Eric Rachlin. See html