Polymath6: A is to B as C is to ???

The Polymath experiment is still very much in its infancy, with the result that we still have only a rather hazy idea of what the advantages and disadvantages are of open online multiple collaboration. It is easy to think of potential advantages and disadvantages, but the more actual experience we can draw on, the more we will get a feel for which of these plausible speculations are correct.

In my last post but one I outlined a few thoughts about the cap-set problem. Although this wasn’t meant as the beginning of a Polymath project (as I said in the post, it was modelled more on Gil Kalai’s notion of an “open discussion”), it had something in common with such projects, to the point where one of the worries people have about the Polymath approach actually occurred: I suggested a line of attack that was, unbeknownst to me, actively being pursued by somebody.

I do not know exactly what calculations went on in the minds of Michael Bateman and Nets Katz when they decided to go public with their ideas before they had fully worked them out, but my guess is that they wanted to establish that they had got there first. This they did by the very effective method of explaining that the simple observations that I had made were much weaker than what they already knew how to prove. As it happens, the story has had a happy ending for them, since they managed, soon after posting their comments, to push their ideas to the point where they have obtained the first improvement to the Roth/Meshulam bound, something that many people have wanted to do for a long time.

It’s possible therefore that the quasi-Polymathematical post of mine was actually beneficial to Bateman and Katz, providing them with just enough fear (almost certainly unwarranted, I should add) of getting scooped that they were stimulated into pushing their already excellent ideas through to their conclusion. I’d need to know their side of the story a bit better to know whether that really is the case, but at any rate I think it is fair to say that the outcome has been positive.

We now find ourselves in a very interesting situation. The argument of Bateman and Katz improves the Roth/Meshulam bound by a factor of n^\epsilon for a very small absolute constant \epsilon>0. That is, the density they need is 1/n^{1+\epsilon} instead of the old 1/n. Meanwhile, over in the \mathbb{Z}_N context, there has been much excitement over the recent result of Sanders that improves the bound for Roth’s theorem to a density of (\log\log N)^5/\log N. We should think of N as the equivalent of 3^n in the \mathbb{F}_3^n case, so \log N is the equivalent of n. Thus, the Bateman-Katz bound would, if one could carry it over to \mathbb{Z}_N, give a bound of 1/(\log N)^{1+\epsilon}, and that gets people very excited because it would prove that every set of integers A\subset\mathbb{N} such that \sum_{n\in A}n^{-1}=\infty contains an arithmetic progression of length 3, which is the first non-trivial case of a very famous conjecture of Erdős.

Just in case the meaning of the title of this post is not obvious to everybody by this point, let me spell it out: let A be the the Roth/Meshulam proof, let B be the proof of Bateman and Katz and let C be Sanders’s proof. It is tempting to speculate that there exists a proof D that stands in relation to C as B does to A, allowing an improvement to Sanders’s bound by a factor of (\log N)^\epsilon.

The most interesting question concerning this speculation is of course whether it is correct. But the second most interesting question (well, perhaps I don’t really want to claim that, but any rate, an interesting question) is what is going to happen now. There are several people around with a strong interest in these problems and experience with similar questions. Should there now be a race, with all the glory going to the person, or small team of people, who can most quickly digest the existing arguments and find the right combination of them?

That would be the obvious non-Polymath way of doing things. The obvious Polymath way of doing things would be to do things the Polymath way — obviously. That is, everybody who wants to should participate in an attempt to understand exactly what is going on in the two recent papers (as well as certain other papers that have contributed to making these breakthroughs possible, notably papers of Croot and Sisask and of Katz and Koester) and to work out whether we do now have enough ingredients to break the 1/\log N barrier in Roth’s theorem.

The main purpose of this post is to say that a Polymath project of this kind is indeed going to start. This comes after quite a bit of emailing in the background, which has established that all of Sanders, Bateman and Katz are happy to do things this way (where I think “happy” means genuinely happy rather than agreeing-after-twisting-of-arm happy), as well as several other interested parties.

One of our concerns is that nobody who might be interested in participating should be excluded, so this is the moment where the private emailing stops and the conversation can take place in public. At the moment, the focus is on trying to understand the papers of Sanders and of Bateman and Katz, though there has been a small amount of assessment of what might need to be done, which I’ll discuss in a moment. An early priority is to set up a wiki, where people can add articles that contribute to this understanding. (I’m hoping, for instance, that people may write articles in which they try to explain what the ideas behind the various papers are: it would be interesting to have several people writing such articles, each with slightly different perspectives, even if the articles overlapped.) I have created a page for this purpose and added links to the articles of Bateman-Katz and Sanders. It can be reached by clicking here or by following a link in the sidebar of this blog.

I find myself unable to do any more work for a few hours, so I’ll post this as it is and extend the post later. I have created a very brief initial post on the Polymath blog. I suggest that mathematical comments should go there, whereas comments on the general Polymath issues discussed above would be more appropriate here. I’m hoping that people will regard the mathematical discussion as being in a preliminary stage, where the main purpose is to help as many people as possible get to the point where it is feasible for them to think about the problem, rather than to forge out ahead. However, it seems difficult (and not obviously right) to dictate this.

18 Responses to “Polymath6: A is to B as C is to ???”

  1. David Says:

    I can only commend everyone’s spirit of cooperation. It’s what I feel maths should really be about – moving forward together to solve old problems and discover new things.

  2. Sakura-chan Says:

    What happened to the EDP?

    • gowers Says:

      It just sort of gradually wound down as the participants lost energy. However, I have it on my to-do list to write a survey paper (with the assistance of anyone who wishes to join me) summarizing what we learnt during the project. Also, I don’t rule out returning to it at a future date, as I still feel as though there are some ideas to explore.

    • Gil Kalai Says:

      Maybe we should have a reunion EDP post a few months from now. (I had planned to write something myself but it goes slowly.)

    • gowers Says:

      Something like that would suit me very well.

    • Klas Markström Says:

      It is still an interesting problem, but for me other things have been taking up my time and energy.

      A survey paper sounds like a nice idea, and thinking about it might wake some new ideas as well.

  3. Polymath6 « Euclidean Ramsey Theory Says:

    […] has started. It is about improving the bounds for Roth’s theorem. See this post. Also there is a page on the polymath blog here. There is also a wiki here. Also see this post and […]

  4. Polymath6: thoughts on empirically testing the finite-field heuristic for very low ‘n’ « Paul Delhanty Says:

    […] Gowers has now officially launched Polymath 6. The scope of the project is well described in his blog post, so I won’t repeat it here. […]

  5. valuevar Says:

    Do you think it would be useful to have polymath sessions in the flesh from time to time? Of course, not everybody could be present for any given session, but it wouldn’t be half bad if at least active contributors in the same geographical area could get together for a weekend full of nothing but work.

    • gowers Says:

      That’s an important point that we’ve discussed. Given that several people are passing through Cambridge in the near future, there are obvious opportunities for that kind of interaction. Indeed, Ben Green, Tom Sanders and I have already had various discussions.

      I get the impression that the way of working for this project may be a bit different from how it has been with previous projects. Somehow a rapid throwing out of half-formed ideas seems a bit less helpful: we all have a fairly clear idea of the kind of thing we are trying to do, and the difficulties involve getting to grips with some issues that may be technical and may be more fundamental. It could be that the right way to proceed is for people to do a bit more private thinking, and to write as much expository material as we can, with the aim of all helping each other to understand the two key papers and the other techniques that are around.

    • Gil Kalai Says:

      It seems that it will not be realistic to hope that we are going to observe a massive open collaboration even in the modest meaning of the word “massive” from polymath1-5. However, if a larger set of observers can watch (and perhaps even make remarks occasionally) over the polymath blog how a smaller (hopefully non empty) set of researchers openly trying to solve the problem this can be a useful experience as well. (And as mentioned above off-line conventional collaboration by the small group of involved researchers can be useful.)

      Of course, it seems perfectly reasonable if other mathematicians will study separately this problem on their owns in a more conventional matter.

      One advantage of having such an open research is that it makes it easier for the community to understand the issues, ideas and proofs.

      There is a famous example of Ian Agol who openly studied and eventually proved over his blog the famous tameness conjecture. Proofs of the conjecture were reached also by Suhyoung Choi who had a different approach and used some ingredients of Agol’s open ongoing work to complete it, and by Dan Calegari and David Gabai who independently worked on the conjecture and solved it.
      (I hope I tell the story right.)

  6. Gil Kalai Says:

    Bateman and Katz’s improvement on the cap set problem upper bounds is certainly exciting and moving to pushing the exponents of log n beyond 1 is certainly a natural thing to try. I would be very happy if we will also keep discussing lower bounds. Lev’s startling example, the remarks regarding it and also Tim’s example are all very interesting and so is Seva Lev’s suggestion in this remark https://gowers.wordpress.com/2011/01/18/more-on-the-cap-set-problem/#comment-10601 is also very interesting. So I propose that perhaps beside the pointed effort to examine the possibility of improving Roth we will also discuss or examine possible examples to the cap set problem (and also to Roth). (E.g. is Seva Lev’s example transferabble to the Roth problem?)

  7. gowers Says:

    Over on the Polymath blog I have put a new post in which I discuss whether it is possible to modify Bourgain’s Bohr-sets argument to obtain a 1/\log N bound instead of using Sanders’s rather different methods. I am far from sure that there is any future in this line of attack, but I can’t rule it out either, and if it worked, it would mesh better with the Bateman-Katz arguments.

  8. Anonymous Says:

    Sorry to leave you a comment unrelated to this post (or tangentially, at best).

    Already in your book “Mathematics: A Very Short Introduction” you noted the “myth of music and maths”, and recently I stumbled with an extended opinion of yours published in the “The Independent”.

    My question is: does the fact that there exists a branch of Mathematics dealing with the abstract (and some concrete) aspects of music alters in some way your perceptions about the issue?

  9. Set is not the only version of Set « Quomodocumque Says:

    […] this problem in a long time is the recent paper of Bateman and Katz.  Some discussion by Gowers of the relation between this paper and the polymath project. GA_googleAddAttr("AdOpt", "1"); GA_googleAddAttr("Origin", "other"); […]

  10. Terence Tao Says:

    It looks like Thomas Bloom has now achieved some version of Polymath6, although still a little shy of the 1/log N barrier: http://arxiv.org/abs/1405.5800

  11. Polymath 10 Emergency Post 5: The Erdos-Szemeredi Sunflower Conjecture is Now Proven. | Combinatorics and more Says:

    […] cap set development may reflect also on bounds for Roth’s theorem, (little in the spirit of polymath6: “A is to B as C is to ???” ) with  some naive thoughts about it. I still hope for some serious discussion about such a […]

  12. To cheer you up in difficult times 7: Bloom and Sisask just broke the logarithm barrier for Roth’s theorem! | Combinatorics and more Says:

    […] Gowers proposed in 2011 polymath6  to try to break the logarithmic barrier for Roth based on the Bateman-Katz breakthrough. (Here is […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: