Skip to content

Goin’ up, down, all around, it’s like a see saw*

July 16, 2014

This is my last research life-story (at least for now), possibly concluding this project (though you are all very welcomed to share more as long as this blog lives). My main hope was to give legitimacy to all of us to acknowledge and discuss our uncomfortable feelings and the “non-scientific” challenges of our careers. My experience with myself and others is that many of these neuroses are quite universal. And they are not necessarily correlated with success, which sometimes only adds internal pressure. Paraphrasing what Russell Impagliazzo told me the first time we met (years ago): we really are competing with ourselves, and this is a hopeless competition (I’m sure he said it better). As for myself, I feel that I learned how to enjoy our profession much more over the years (mainly through becoming a little less childish with time). Still, at times, I do feel inapt. Such a period is the topic of my last story.

During my last postdoc year, we had our first child. This was a wonderful event that I had been craving for years. But it was also very demanding. My son was colicky and we were inexperienced and mostly alone in the U.S. In addition we had three house moves, one of which was back to Israel (a move that was surprisingly non-smooth). I was very content with putting my young family at the center and I realized that this is a period that will not repeat itself and should be cherished (turns out that with kids, many periods are like that). I also understood that I cannot expect to do too much research at this period. There was nothing concrete I was worried about: I had just landed my dream position at Weizmann, I wasn’t worried about getting tenure, and I already had many results that I was very proud of (including one with Irit Dinur on PCPs that was quite recent). I could allow myself to take it easy, but my ego was not ready for that. With time, internal pressure accumulated. “Is this it? Did my creativity dry up? Is it downhill from now on?”

At the end of that year at Weizmann (with my son being just a bit over a year), I headed with my family to a summer trip to Berkeley (to work with Luca Trevisan and Irit Dinur) and to Cambridge (to Work with Salil Vadhan). I decided to invest all of the effort in problems related to RL vs. L and felt that this is a test for me. If I’ll fail, then I will scale down my expectation of myself. With this shaky (and so very silly) state of mind, I came to a complexity-theory workshop that started the trip. Though my talk about the work with Irit was very well received, I felt quite depressed. It felt like everyone have been doing these wonderful research and only I was idle. I especially remember one of these talks, with a speaker (who I knew to be very nice) that had an over-confident demeanor. Such individuals always put me off, but at this strangely vulnerable state of mind, it was a challenge to keep the tears inside.

The summer continued quite differently. Spending time with wonderful friends (who happen to be brilliant researchers), having a lot of time for vacationing with my family (thanks to Luca’s great life balance), and ending up with a result that exceeded all of my hopes (undirected connectivity in log-space). I remember very vividly the exact moment when two ideas that I had for many years suddenly clicked in an exciting new way. I remember pacing back and forth in our hotel room, going over the proof that then only existed in my mind. Could it be true? Surely not! But where could the bug be hiding? I remember going out to find a store that would have a notepad and pen for me to start writing the proof down and the slowly growing confidence that came from writing it down and every session of verification (Luca, Irit, Salil, …). And most of all, I remember all of the colleagues being happy with me and for me.

I am not sure if there is a lesson to be learned here. Perhaps, don’t believe everything you are feeling. Or at least – if you are neurotic, you are not the only one here.

* title inspired by Aretha .

FOCS 2014 Accepted papers list is online

July 15, 2014

The accepted papers list for FOCS 2014 is now posted online.

I am always amazed by the depth and breadth of works in the TCS community, and this FOCS is no exception. Whether you are a physicist interested in the possibility of general “area law” governing entanglement between different parts of systems, a geometer interested in Gromov’s topological notion of expanders, an optimization expert interested in the latest and greatest in interior point methods, a game theorist interested in Karlin’s longstanding conjecture on convergence of fictitious play, a complexity theorist interested in the latest efforts to separate the determinant from the permanent, or simply a dog owner or triangle lover, you will find something of interest in the conference. And of course FOCS is not just a collection of unrelated papers. A quantum computing expert would want to check the paper on topological expanders, as similar concepts have arose in the context of topological approaches to quantum error correction. An optimization expert might want to understand the convergence of “fictitious play” which is a very natural algorithm for solving linear programs, and of course since STOC 2014 we all know  that circuit lower bounds are tightly connected to improving the exponents of algorithms for combinatorial problems. This is just a taste and I could have chosen many other such examples, all strengthening Avi Wigderson’s point why we should all go to talks in areas other than our own.

I was also amazed by the effort reviewers and program committee members have put in the selection process. Conference reviewing sometimes get a bad reputation as being superficial. I did not find this to be the case at all. People have invested an amazing amount of work reading the papers, checking proofs, chasing down references, verifying technical points with the authors and other experts, and generally doing the best job they can to have an informed selection process and assemble the best program we can for the TCS community. I am sure we made mistakes, and the final program, as a product of a committee, is not fully consistent with any particular PC member’s taste, including my own. In particular, there were many submissions that some of us personally found novel and interesting, but were not included in the final program. But I do feel good about the process and believe that while some of our decisions may have been wrong, they were not wrong because we were superficial or lazy or cut corners due to the time pressure. Many times during this process I asked the PC members to go above and beyond what is typically expected, and they have more than risen to this challenge, often making heroic efforts to understand very complex (and sometimes not so greatly written) papers, and trying to get to the bottom of any misunderstanding. I am deeply grateful to them all.

Finally, some statistics. We accepted 70 papers, which is about 26% of the 268-273 submissions (depending on whether you count withdrawn ones). Aside from 9 submissions that were judged to be out of scope and received minimal reviewing, on average each submission had 3.3 reviews and 11.7 comments (including both comments by the PC and short comments/opinions by outside experts that were solicited in addition to the reviews.) Of course these numbers varied greatly based on how much attention and investigation we felt each submission needed and there was also extensive discussion on some of the papers during our two long days of the physical PC meeting. Finally, a very personal statistic for me  is that there are about 2800 emails in my “FOCS14″ folder.  As many past chairs told me, the best thing about this job is that you only do it once…

 

 

 

Collaboration, competition, and competition within collaboration

July 1, 2014

Another instalment on my research-life stories.

—————

The Talmud says: “competition/envy among scholars increases wisdom” (kinat sofrim tarbe chochma). Good or bad, competition is here to stay. Nevertheless, one of the strengths of our community is in its collaborative nature. This is good for science, but in my eyes also makes our life so much better. A recent example is a research project with Guy Rothblum. For a few weeks, we met quite regularly and every meeting went more or less as follows: First, we would go over the solution from the previous meeting and find a bug. Then we would work together on a new and improved solution. This sounds frustrating (and would probably have been frustrating if I worked alone), but instead it was a great joy. We got to solve this problem again and again, and in the process enjoy each other’s creativity and company. Unfortunately, our current solution seems quite robust, so our fun ritual ended.

My best example for turning competition into collaboration is in my long-term collaboration with Salil Vadhan. It started when Ran Raz and I had a modest result on Randomness Extractors (following the breakthrough work of Luca Trevisan). We then learned that Salil had the same result, and already managed to write it down. Salil invited us to join (and I’m sure he was a bit sad to lose his first single-authored paper), on the other hand, Ran and I decided to decline and give up on the result altogether (and I was sad to lose a paper at this early stage of my career). In retrospect, losing that result would have been quite inconsequential, and similarly for Salil. But what did turn out to be extremely significant was what happened next. The three of us started collaborating together, leading to a stronger paper and then an additional collaboration, and before long Salil and I established not only a long-term research collaboration but also a great friendship. The unfortunate accident turned out to be most fortunate after all! Not all collaborations end up so fruitful, but I almost never regretted a collaboration (DBLP gives me 74 coauthors so this is a large sample). I hope that the set of collaborators that regret working with me is equally small.

So let’s all choose collaboration over competition and happily ride into the sunset. Right? Well, not so fast. Collaboration and competition are not mutually exclusive. Turns out, we cannot shut down our egos even when we enter a collaboration. While I strongly believe that the contributions to a collaboration cannot be attributed to any one of the contributors, we all like to feel that we contributed our fair share and that we demonstrated our worth (to others and more importantly to ourselves). An over-competitive collaboration can be destructive, but in moderation it could indeed be that competition among scholars does increase wisdom.

Woos and boos: my research talks

June 24, 2014

Coming back to the research-life stories project I intend to write a few (three that currently come to mind) more stories of my own, hoping that they will inspire more stories by others.

—————

My first research project progressed very quickly. A few months after I started working with Moni, I found myself writing my first paper (having only read very few papers before), and then preparing my first research talk (having attended very few research talks before). My first talk was at Weizmann, where I managed to utterly confuse some of the most brilliant scientists of our generation. During the talk I apologized and Adi Shamir said something to the effect of “it’s not you it’s us.” After the talk, Oded Goldreich made sure I will not be misled by Adi’s gallantry :-) Indeed, it wasn’t them – it was all me. I learnt some valuable lessons about research talks (for example that the intuitions that lead your research may be irrelevant and even confusing when presenting it). It was also the first of many opportunities for Moni to (try to) teach me to never apologize. Unfortunately, even that Moni is completely correct, the temptation is often just too strong for me to resist.

Soon after, I was getting ready for a trip to my first conference and a few seminar talks I managed to schedule. I was quite terrified. Afraid to mess up the talks, afraid to expose my ignorance and even anxious about the practical aspects of travel (it was only the third time I left Israel, and the first time I did it by myself).

On the night before my early morning flight, I get an email addressed to Moni, to me and to a large collection of dignitaries (for example, David Harel, whose only fault was being a past department chair). The email from Professor X went something like “I saw the abstract of your coming talk at MIT. You claim to give the first construction of Z. This is arguably a big fat lie as Professor Y and I already did it in our paper. The only honorable solution on your part is hara-kiri“. I was horrified. I didn’t know their paper and his claim could easily have been correct (it couldn’t have but I wasn’t fully aware at the time of Moni’s encyclopedic knowledge). It was too late to call Moni and I was afraid he will not see the email before my flight. My deck of slides was printed and my first talk was the morning after I land. Disgrace was imminent!

Fortunately, very soon after, Moni replied. I think that Professor X never answered Moni’s email and years later when he interviewed me for a faculty position he have shown no sign of remembering this incident. Professor Y accepted Moni’s explanation (in a private email if I remember correctly) but still managed to squeeze in a couple of rather aggressive questions during my conference talk. By then I was fortunately prepared and calm (and answered: “are you the first to do Z?” with an innocent “yes”).

My second talk during the trip was in MIT. Between the regulars and the visitors, my audience included half of my reference list. I was in awe. The talk was extremely vibrant, with many questions from the audience and if my memory serves, especially from Leonid Levin. I was ecstatic, and I did not mind at all that Adi and Oded (who were just starting a sabbatical in MIT) were taking on many of the questions directed at me. And why should I mind? Here are so many of my idols vividly debating my work! Who am I to disturb them? I only realized it might have been unusual when Michael Luby (giving a talk the day after) answered the first question he got with something like “unlike yesterday’s speaker, I’d like to talk more than five minutes … .“ Still, over dinner, Oded promoted the idea that it may be better if a talk is given by someone other than the authors. So I felt somewhat vindicated.

Throughout the years I gave many more talks, some praised, some scowled, and some both (sometimes even by the same person). Giving a bad talk can be painful (and when you are young it sometimes feels like the end of your career). I vividly remember how the criticism over my first practice job-talk paralyzing me for almost all of the time I had before the first interview (till Moni gave me a few simple comments that dramatically improved the talk). I am still beating myself for not customizing my ICM (International Congress of Mathematics) talk to the non-cs audience. On the other hand, giving a good talk can be quite empowering. One of the sweetest comments that I remember came from Avi Wigderson who told me after a survey talk on RL vs. L that I left the audience no choice but to work on the problem. Like many other things in life, the more you invest in preparing a talk the more you get out of it. While I enjoy giving talks a lot, it is very hard to recreate the rush that comes from overcoming fear in those early talks of my career. I doubt that I sufficiently appreciated this rush at the time (a Joni Mitchell song comes to mind).

Independent conferences: the second-worst solution

June 10, 2014

The steering committee of the Conference on Computational Complexity has decided to become independent of IEEE. The Symposium of Computational Geometry is considering leaving ACM for similar reasons.

I completely understand the reasons, and applaud the steering committees in both cases for having a thoughtful, deliberate, and transparent process. Indeed, I have signed the letter of support for CCC. But, I am not happy about this outcome. I think that having our conferences under an umbrella such as ACM or IEEE that unities much of Computer Science is a positive thing, regardless of practical issues such as bank accounts, insurance, hotel deals etc.. (that thankfully I understand very little about) or even issues such as “prestige” and library subscriptions. I am afraid that an administrative isolation of a sub-area might end up contributing to an intellectual isolation as well. For example, while the idea might sound appealing, I think it is good that we don’t typically have “department of theoretical computer science” (let alone a department of computational complexity or computational geometry). It is important for us to interact with other computer scientists, if only so that we can occasionally torture them with an equation-filled colloquium talk :)

As I said in the past, I wish that our professional societies would behave more like their mission statements and less like for-profit publishers, and so conferences would not feel compelled to leave them. Given the hundreds of votes in the CCC and SoCG elections, I can’t help but think that if steering committees and chairs of various SIGs across all Computer Science collaborated together, they could marshal thousands of votes in the ACM elections that would truly change how it operates.

 Update: Please see Paul Beame’s comment below for some of the significant differences between ACM and IEEE.

ICM Survey: Codes with local decoding procedures

April 25, 2014

I have recently completed a survey “Codes with local decoding procedures” and will be giving a talk on this matter in August at the ICM. The survey covers two related families of codes with locality, namely, locally decodable codes that are broadly useful in theoretical computer science and local reconstruction codes that have recently been used in data storage applications to provide a more space efficient alternative to RAID. Below is the abstract of the survey:

Error correcting codes allow senders to add redundancy to messages, encoding bit strings representing messages into longer bit strings called codewords, in a way that the message can still be recovered even if a fraction of the codeword bits are corrupted. In certain settings however the receiver might not be interested in recovering all the message, but rather seek to quickly recover just a few coordinates of it. Codes that allow one to recover individual message coordinates extremely fast (locally), from accessing just a small number of carefully chosen coordinates of a corrupted codeword are said to admit a local decoding procedure. Such codes have recently played an important role in several areas of theoretical computer science and have also been used in practice to provide reliability in large distributed storage systems. We survey what is known about these codes.

ICM Survey: Sum-of-squares proofs and the quest toward optimal algorithms

April 21, 2014

I have just posted online a new survey “Sum-of-Squares proofs and the quest toward optimal algorithms” co-authored with David Steurer .

The survey discusses two topics I have blogged about before – Khot’s Unique Games Conjecture (UGC) and the Sum-of-Squares (SOS) method – and the connections between them. Both are related to the notion of meta algorithms. While typically we think of an algorithm as tailor-made for a particular problem, there are some generic approaches to design an “algorithmic scheme” or a meta algorithm, that can be applied to a large class of problems. The UGC predicts that a particular such meta-algorithm, which we call in the survey simply the “UGC Meta-Algorithm”, would give in fact optimal approximation guarantees among all efficient algorithms for a large class of problems. This is very exciting from the point of view of complexity theory, not simply because it gives many hardness-of-approximation results in one shot, but because in some sense it gives a complete understanding of what makes problems of certain domains easy or hard.

The SOS method is another meta algorithm that can be applied to the same cases. It is parameterized by a number \ell called its degree. Using a higher degree can give potentially better approximation guarantees, at the expense of longer running time: running the method with degree \ell on input of length n takes n^{O(\ell)} time. The UGC Meta-Algorithm in fact corresponds to the base case (which is degree \ell=2) of the SOS method, and so the UGC predicts that in a great many cases, using constant (even even mildly super-constant) degree larger than 2 will not help get better approximation guarantees. We discuss in the survey a few recent results that, although falling short of refuting the UGC, cast some doubts on this prediction by showing that larger degree can sometimes help in somewhat similar contexts. I will give a talk on the same topic in the mathematical aspects of computer science section  of the 2014 International Congress of Mathematicians to be held in Seoul, South Korea, August 13-21, 2014. As you can see, there will be some great speakers in this session (and ICM in general), and I hope we will see blog posts on some of those surveys as well.

Follow

Get every new post delivered to your Inbox.

Join 72 other followers