Monday, January 11, 2010

Guest Post on ICS 2010 (1 of 3)

Innovations in Computer Science 2010 (post #1)

Guest post by Aaron Sterling

This is the first of three posts about ICS 2010, the much-discussed "concept conference," which took place at the Institute for Theoretical Computer Science (ITCS), Tsinghua University, Beijing, from January 5th-7th. I will provide my impressions in this post and one other, and Rahul Santhanam plans to contribute something as well.

First, I need to say that this was the best-run conference I have ever attended, and one of the best-organized events of any kind that I have ever participated in. The level of financial support for students and authors, the quality of food and lodging, and the remarkable closing ceremony (which included several music and dance acts, a Kung Fu demonstration, and -- my favorite -- a Face-Off performance) set a high bar for any other conference in the world. Local Arrangements Committee members don't often get mentioned in posts like these, but I believe the entire TCS community owes a debt of gratitude not just to PC Chair Andrew Yao, but also to Local Arrangements Chair Amy Yuexuan Wang, Conference Secretary Yuying Chang, and to everyone else who made this event happen. This feels like a turning point in the history of the field.

In the US, I have often gotten the impression that computer science departments and funding sources consider TCS to be of secondary importance. What a difference in Beijing! As a silly-yet-telling example, Sanjeev Arora told me that, for a conference in 2009, ITCS printed a sign in which the phrase "Theoretical Computer Science" appeared in the largest-size font ever. I believe the investment in theory on the part of the Chinese government and academia, contrasted to the malaise of departments in the United States, speaks volumes about the future, unless the United States changes direction significantly. I'll leave that topic to be discussed on a political blog, though. Suffice it to say, I think everyone was pleased to be treated like a first-class scientist, instead of like someone doing "impractical" things that are less worthy of support.

Perhaps the highlight of the technical program was the "derivatives paper," already covered at length by Richard Lipton and other bloggers, so I won't discuss it here. Many of the accepted papers were in algorithmic game theory, and I will limit myself to mentioning the two papers in that area I found the most exciting. These are "Bounding Rationality by Discounting Time" by Fortnow and Santhanam, and "Game Theory with Costly Computation: Formulation and Application to Protocol Security" by Halpern and Pass. Essentially, Halpern and Pass define a class of games with complexity functions attached, so it is possible to reason about concepts like equilibrium with respect to a particular measure of complexity. The Fortnow/Santhanam model embeds into this approach, as it considers one particular type of complexity function. On the other hand, the complexity function defined in Fortnow/Santhanam seems particularly natural, and they are able to obtain more specific results than Halpern/Pass, because they start with a less generalized model.

The conference started off with a bang: Benny Applebaum gave an excellent talk about cryptography obtained by using only local computation. This was "Cryptography by Cellular Automata or How Fast Can Complexity Emerge in Nature?" co-authored with Ishai and Kushilevitz. They constructed, for example, one-way functions with one step of cellular automata (i.e., after one step, it is computationally hard to invert the state of the system to the original state). As cellular automata can only communicate with their immediate neighbors, this has bearing on the parallel complexity of cryptography. One point that came up in discussion is that, unlike one-way functions, document signatures cannot be obtained by local computation only, because of the need to make global change to the output if a single bit of the input is changed.

The "Best Impromptu" Award goes to Avrim Blum, who, on three hours' notice, gave one of the most stimulating talks of the conference when he presented "A New Approach to Strongly Polynomial Linear Programming" by Barasz and Vempala, after the authors had a problem with their trip. The Barasz/Vempala concept is a hybrid of the Simplex Algorithm and the Interior Point Method for solving LP's. Rather than just trace the edges, or just go through the interior of the polytope, they take the weighted average of the "useful" edges near the current location, and follow the obtained "averaged" line until they hit another face in the polytope. It is unknown in general whether their algorithm runs in polynomial time, but it seems very interesting, because they have shown that, for each case for which Simplex runs in exponential time, their algorithm can solve that "hard case" in polynomial time. This is because their solution method is invariant under affine transformations of the problem statement, so it is robust even when the angles of the polytope are narrow, i.e., the constraints are very close to one another.

I will conclude this post by mentioning Bernard Chazelle's "Analytical Tools for Natural Algorithms." (Please see a previous guest post of mine, and comment 3 of that post by Chazelle, for some background.) His main philosophical message was: "Use algorithms to analyze algorithms" -- meaning that if one is trying to analyze the behavior of a nonlinear multi-agent system like ABC...Dx, where A,B,C, ... ,D are matrices whose identity depends on time and some kind of feedback loop, it is not helpful to consider the problem "just mathematically," by analyzing the operator ABC...D independent of x. Rather, one should consider the problem in the form A(B(C(Dx))), and design an algorithm to reason about this nested behavior. That algorithm can then (hopefully) be tweaked to prove similar results about related nonlinear multi-agent systems. To quote from his paper: "Theorems often have proofs that look like algorithms. But theorems are hard to generalize whereas algorithms are easy to modify. Therefore, if a complex system is too ill-structured to satisfy the requirements of a specific theorem, why not algorithmicize its proof and retool it as a suitable analytical device?"

In my next post, I'll sketch results from a few more papers, try to give some flavor of the discussion session at the end of the conference, and offer a few suggestions for the future. My apologies in advance to all the authors whose work I will be leaving out. Several attendees commented how accessible and well-presented the talks were -- and I noticed this too. (I think this was due in large part to the "call for concepts." Certainly when I prepared my own presentation, I felt motivated to communicate "high" concepts as well as I could, and I felt less pressure to include the Mandatory Scary Formula Slide(tm) to demonstrate my ability to perform rigorous research.) In any case, there is far more great material than I could possibly cover in two posts -- which is a very good problem for a new conference to have!


  1. Excellent post, Aaron! I look forward to reading the rest.

  2. This feels like a turning point in the history of the field.

    Not to take anything away from the organizing committee, but I think it's just a sign that they had lots of money to spend -- not a turning point for our field.

  3. I fully support the Chinese spending as much money as possible on the least applicable parts of computer science (computational complexity, algorithmic game theory, etc.).

    That has to be the least threatening thing they could possibly be spending their treasure trove of US currency on. Maybe theory will save us all, after all...

  4. "This feels like a turning point in the history of the field."

    no way, really?

    In any case, I was wildly amused to read.

    "...closing ceremony (which included several music and dance acts, a Kung Fu demonstration, and -- my favorite -- a Face-Off performance) set a high bar for any other conference in the world."

    I have no idea if a Kung Fu demonstration or even Karate or any other martial arts performance combined with the wildest acrobatic performances on earth or even the funniest movie clips or the fastest 100 meter runners on the planet, contribute to "higher bars" for a TCS conference. Seems to be a new selling point upon which conferences can be judged by ? I really hope that wont be the case.

  5. Don't you think it would be somewhat rude of Aaron to report on the conference and make fun of the entertainment? Maybe he was just trying to be polite. I doubt he wants Kung Fu performances from now on ...

  6. I agree, the post sounded like someone is in need of getting a job there.

    This feels like a turning point in the history of the field.

    It seems Aaron is not familiar with the quote, all that shines is not gold .

  7. The author absolutely do not understand China. Can you find another person in China who can organize conferences like this? And then compare with US!

  8. Wow. Okay, let's back up a bit folks.

    In one corner, conferences (including STOC and FOCS) are talking about eliminating paper proceedings to save money, or have already eliminated them. Many other cost-cutting measures are being considered. In the other corner, you've got a conference, which, in its first year, has a 500-page printed proceedings, an impressive technical program, and also provided the attendees a variety of unnecessary extras, like the Kung Fu performance. (In fact, you'll see in my next post that I suggest investing the resources of the closing performance elsewhere, but that's another point.)

    Institutionally, contrast a five-year-old research group that is recruiting some of the most distinguished visitors in the field, to the University of California system, which is under siege.

    If you are unable to see the (very obvious) point that this has significant sociological ramifications for the future of computer science internationally, then frankly I have to wonder whether you are allowing the grinding of some political axe to get in the way of your ability to assess what is good (and bad) for science.

    The fact that there was a nice show is not the point. The fact that there was extensive funding and superb organization -- especially contrasted to the defensive situation of TCS in the United States -- is what makes me believe the field is changing.

    I won't be surprised if some more anonymous snarkers decide to impugn my character further, but I won't respond again.

  9. If the martial arts performances were not such a big deal -- well -- then don't dwell on them by expanding on how dazzling they were in your description. Otherwise, you run the danger of being misinterpreted.

    Also, don't get defensive, reading all comments, I don't think any commenter impugned on anything.

    The punchline here is that with enough money and government funding, anything can be done. This is certainly not big news.

    Also, don't exaggerate the defensive situation of TCS in the United States .

    500-page printed proceedings(!), poor trees! The idea in the U.S. is to Go Green all the way, on all possible aspects. That's called E-V-o-L-u-T-i-O-n.

  10. Doesn't anybody mind that China is a country with very little respect for human rights and serious limitations on personal freedom. I thought we scientists were supposed to care about these things...

    Free Tibet!

  11. US has very little respect for human rights as well. US pretends to respect human rights but goes on abusing - Iraq, Afghan, Guantanamo Bay, Abu Gharib, ... the list is endless

  12. Is this kindergarten or what?

    The Kung-fu is not the issue! Aaron gave a nice and interesting overview, and these ad hominem posts are disgraceful. Seems people just need some target to aim their aggression to. If you have something meaningful to say say it decently and then you won't need to write ANONIMOUSLY.

    Otherwise, find some other blog to write your s**t.

    Noam Livne (nonanonimous)

  13. I feel so funny that most of the comments on focus on the second paragraph of the post. People seem to miss the rest of the post.

    Accept it or not, it's really well-known facts for grad students like me that it's really hard to get an academic position recently if one works on theory. The truth is most of us go to industry or business after getting our PhD. The same situation also applies for mathematicians.

    It's a shame if someone after his PhD stays jobless. Even a garbage collector gets $80,000 in New York City. Don't forget the story of Dr. Douglas C. Prasher, an American molecular biologist who becomes a bus driver. I guess it's also called E-V-o-L-u-T-i-O-n then?

  14. Guys, the last couple of posts amount to kids' talk. One anon was advocating to free Tibet, the other anon was retaliating with his comments. Where did we end up ? Can't you show a little bit more respect ? Remember you are all visitors on this blog.

    While I do agree that "Free Tibet" and "Go Green" should be advocated, I would not do these things on this blog as that would be the same thing as advocating "Martial Arts Performances" at places that aren't really meant for them.

    The point here is very, very simple. China has a lot of money and it will think strategically of how to disperse it. It does sound very surreal that a 5 year research institute can form a PC of that caliber in such short time. But given the support of the communist party and the funding, sky's the limit as to what you can accomplish.

    DTML, while garbage collectors in china might be billionaires, they certainly do not earn $80,000 in New York. I'm a former NYer. More false stats is always welcome, as it warms up the cold atmosphere or climate.

  15. Are they going to youtube some of the lectures?

    I am from China. KungFu is nice but not necessary. That is just the way they do things in China. China is changing rapidly, in an unknown direction though.

  16. Very nice summary, thankyou for writing it, Aaron. Agree with Noam - the cowardly anonymous posters picking irrelevant (indeed, imaginary) holes in the post should grow up.

    Michael Nielsen

  17. I am anon #2. Just to clarify: when you wrote "this feels like a turning point in the history of the field", I thought you meant "all conferences from now on will have to live up to these standards" (something that I thought was ridiculous). I now realize that what you actually meant was "the strength of TCS is shifting to China", a statement I agree with.

    Sorry for any misunderstanding, on both ends.

  18. Doesn't anybody mind that China is a country with very little respect for human rights and serious limitations on personal freedom.

    No, not really. It's not my business to criticize Chinese politics. Certainly not in a complexity blog.

    I thought we scientists were supposed to care about these things...

    On the contrary. A lot of great figures, especially in mathematics, did not care about politics. And to the extent that they did care, they lost something from their dedication to science.

    Free Tibet!

    Why? China has serious claims about Tibet, and we should see both sides of the story.

    If you have something meaningful to say say it decently and then you won't need to write ANONIMOUSLY.

    I disagree. Anonimity is a legitimate part of the internet, and your criticism itself is not entirely "decent".

  19. i am anon #2 and i still disagree with your point. someone else claimed before that he was me but that's not true.

  20. not quite chinese2:43 PM, January 14, 2010

    China has serious claims about Tibet, and we should see both sides of the story.

    Would you care to elaborate? I have not, to date, heard one serious argument why China should have control of Tibet, and I have heard many people try to make such an argument. I am fairly confident that no reasonable person would find any legitimate claim of China over Tibet.

    (One line of arguments that I am not counting here is those of pragmatic nature -- saying that at the current situation, with so many Han Chinese already settled in Tibet, removing them would do more harm then good; or that breaking Tibet off will cause the entire country to seperate, which might hurt more people than it helps; but these are not "good claims for Tibet", they're just pragmatic arguments, and also are false in my estimate).

  21. yeah, China has all the right in the world to take over Tibet!

    There are a gazillion reasons for this.

    1) China has the right to claim Tibet, because it's China.

    2) China has the right to claim Tibet, because it's China ?

    3) China has the right to claim Tibet, because it's (un)just ?

    4) China has the right to claim Tibet, because it's human rights (un)friendly ?

    5) China has the right to claim Tibet, because it's (un)diplomatic.

    6) China has the right to claim Tibet, because it wants to.

    7) China has the right to claim Tibet, because if it wanted to claim the U.S. it would do the same thing.

    8) One day China will claim The U.S., it's just a matter of time.

    9) China is the greatest and biggest superpower in the world.

    10) China is dominant in almost all sciences and sports and will dominate and rule everything in 10 years from now.

  22. Why is that a spectacular theater when speaking about China doesn't surprise me at all ...