My thoughts on CCP

The Division of Computational Physics of the American Physical Society has now its own blog, brought to you by its very own executive committee, of which I am presently a “member-at-large” (soon to be apprehended and brought into custody, I guess). The latest post advertises the upcoming conference on computational physics (CCP 2011), to be held in Gatlinburg (TN), October 30th – November 3rd, 2011.
Now, what exactly is this conference all about ?

I attended CCP 2010 in Trondheim, Norway, where I was invited to give a talk, which was scheduled in a session including two other talks, on subjects essentially unrelated to that of my own — I mean, I suppose all of them would fall within “condensed matter physics” (broadly construed), but the science was very different, the only true common feature being that we all use computers to do our calculations. Not surprisingly, very few of those in the audience who came for my talk, stayed after I was finished; on the other hand, a bunch of other people came in the room to listen to the next one. My impression, from looking at all other sessions, is that mine was not at all an “odd” one.
I have to confess, I do not think that I would be using my own travel money to attend a conference like that, nor would I recommend an advisee to go. I simply do not find compelling enough a reason for sitting through a bunch of talks on random subjects (which, mind you, may be very interesting to others, but not to me), that the results presented were obtained numerically. I suppose I might be interested in a talk on a different subject from that on which I do research, if the methodology utilized in that work could conceivably be useful to me as well — but in my experience that happens exceedingly rarely. As I wrote in this old post, methods that are really effective are generally developed to tackle a specific, narrowly defined problem.

While I do believe that there are valid reasons for having a Division like DCOMP, I still think that at the end of the day we are, and remain, physicists, and that there is nothing intrinsically different about us, just because we happen to be using computers more than our colleagues. I am annoyed by any talk of “third approach”. What we are after, are physical results and conclusions — the fact that they are numerical is only valuable to the extent that they are accurate, reliable and reproducible.
I have stated many times that there is nothing wrong in being interested in methods, and that the development of a novel, innovative methodology constitutes an important scientific contribution, especially if it allows one to gain original theoretical insight into problems that cannot be effectively tackled with existing techniques. Ultimately, however, the main scientific goal of attending a conference (as opposed to a small workshop, which may well be all on methods alone) is to learn about new science; the focus must remain on the results that are obtained, not on the techniques utilized to obtain them (of course, as long as the techniques are believable, robust, extensively tested etc.).

Ensuring some scientific coherence in the various sessions, at a conference like CCP (or maybe through the whole division) may always prove a difficult balancing act. I am afraid, however, that as long as invitations are extended to (and contributed talks solicited of) speakers disseminated throughout the whole discipline, who have accomplished something important using some kind of numerical techniques, the conference may never lose the potpourri character, which in my opinion is not conducive either to learning, nor to increased acceptance of computation as a legitimate investigative tool among skeptics — in fact it may help perpetuate that annoying misconception of numerical people as “technicians”.
I propose a different model, one in which the organizing committee would solicit from groups of scientists proposals for sessions focusing on a specific scientific theme; session speakers would be investigators who are computationally inclined, but whose research work fall squarely within a precisely identifiable sub-field, and who would be talking about the same, well-defined subject.
That would have first of all the effect of making it more likely that the audience find all talks interesting and worth listening to — because they are all interested in that subject. It would also provide some incentive for people to submit contributed abstracts, if they are going to be speaking in some coherent, homogeneous session.
Perhaps most importantly, it would also make it more compelling for others to attend, including some who may not be themselves of the numerical persuasion, but who may be interested in that research subject; ideally, exposing them to concrete, tangible advances made by computation may finally warm them up to it, maybe even to embrace it.

Tags: , ,

13 Responses to “My thoughts on CCP”

  1. Experimentalist Says:

    You are bang on. At the big conferences I never go to sessions on computational physics. I have little interest in the computational methods.

    I also have pet peeves with the way some computational physicists talk/write:
    “I have shown that [e.g. spin ices are made of ice cream] …” No. You modeled something with many assumptions that you did not make clear. “Modeled” is the more accurate word, so it needs to be used.

    The worst is when you can get through the abstract and introduction of an article and are not sure if this was a computational or an experimental work.

    • Massimo Says:

      You modeled something with many assumptions that you did not make clear.

      Whoa, whoa, whoa, hold it right there, bucko…. 🙂
      Sorry, that is not fair. One of the most disturbing misconceptions out there is that “modeling” is something that computational physicists do. That’s nonsense. Models are introduced by theoretical physicists, for reasons that I tried to explain in this post. You may or not be fond of them (I happen to be), but they do serve a legitimate purpose.
      When it comes to studying models, the main difference between computation and analytical approaches, is that numerical results are sometimes believable.
      Making the mistake of adopting a wrong or nonsensical or unphysical model is not a prerogative of computational physicists, nor is it in any way an indictment of computation.

      • Experimentalist Says:

        I’ve read that post before. Apparently it didn’t stick in my mind. I had the distinct impression that computational physicists built the models themselves.

        I don’t have a problem with a first principles calculation or toy models. My main beef is with the opacity of the accuracy of the results.

        As an experimentalist, I probably wouldn’t believe a computational model unless the author was willing to give it a GUI and user instructions for me to test out. For example SRIM or SIMNRA are as believable as the user manuals suggest.

      • Experimentalist Says:

        More importantly, I would stand up and listen if authors of programs I trust said they got such-and-such results with a new program they created.

      • Massimo Says:

        I had the distinct impression that computational physicists built the models themselves.

        No, these are two conceptually distinct stages of the theoretical investigation. The mathematical model comes first, and is fully independent of how you will try to compute its properties (i.e., phase diagram).
        If we want to remain in condensed matter physics, think of the Ising model of ferromagnetism — it was invented way before computers were available to do calculations. Same for many others, like the Hubbard model, or others for which exact analytical solutions were worked out in the 60s (in one dimension) — no need for any computer if you want to know (most) equilibrium thermodynamic properties of those.

        Computation comes into play exclusively when doing the calculation on a specific model is too difficult analytically, and approximate solutions are unreliable because you are away from the perturbative regime (strongly correlated systems). Find me a way to extend Bethe ansatz to two dimensions and I am the first one to say “bye bye” to Monte Carlo. But I am not going to hold my breath for that to happen, and in the end I want to solve the problem, I don’t care how.

    • Schlupp Says:

      Probably you just chose a bad example in “spin ice”, but I do not think that theorists (let alone computational ones, as Massimo explained) are more to blame here than experimentalists: “Spin ice” is a name for a concept, a mostly theoretical one, which was proposed in analogy to water ice long before any “candidate materials” were known.

      So, if you measure ice-cream behaviour in compound A_wB_xC_yD_z and claim to have shown that “spins ice consists of ice cream”, one might just as well reply that you have shown that “A_wB_xC_yD_z consists of ice cream” and have *assumed* A_wB_xC_yD_z to be a realization of the concept “spin ice”.

      • Experimentalist Says:

        Schlupp, I’m trying to read you comment as if it were meant to be funny but it ain’t happening. Has anyone ever told you that you are a ‘fun cop’? i.e. One who goes around policing those that are having fun. Of course it was a bad example. Everyone knows that spin ices are made of bipolar teenage particles to make them so frustrated.

        And I don’t blame theorists, I just don’t want to be seen in public talking to one.

      • Schlupp Says:

        Huh, someone seems to have issues with me contradicting them. I think I will spend the rest of the day being very unhappy indeed.

  2. Experimentalist Says:

    A bit of a tangent…

    I was at CSC in 2009 and the plenary speaker was describing his first day as a professor back in the early days of theoretical chemistry. When he arrived the placard for his office read “Prof. So-and-so, Theoretically a Chemist”.

    Can we give Massimo a similar title? He doesn’t seem to think he is a theoretical physicist but a computational one. “Computationally a Physicist” doesn’t seem to work. How about “Prof. Massimo Boninsegni, Approximately a Physicist”!

    Thank you! I’m here all week!

    • Massimo Says:

      He doesn’t seem to think he is a theoretical physicist but a computational one

      No, you misread me, I think of myself as a theoretical physicist, one who happens to use computers to do the calculations. “Computational physicist” is what I am called by fellow theorists who do not understand how computation works and are too lazy to read about it, and by experimentalists who do not understand how theory works and are too lazy to read about it 😉

    • GMP Says:

      Not that Massimo needs my help here, but I will chime in anyway.
      I am also a “computational scientist” which means I am a theorist who uses both analytical and numerical techniques to solve mathematical models. It’s quite ironic how “real theorists” (pen-and-paper only) think that actually being able to write code makes you less of a scientist [as if the ability to write code means you could not possibly also have the traditional (analytical) theoretical skills. *eyeroll*] In reality, I find numerics quite liberating; if you know what you are doing, first some things can be done way faster and way more accurately than analytically, and you can access a much larger range of physical situations (not just those that lend themselves to analytical approximations). The latter is particularly important if you ever want to achieve quantitative agreement with experiments, because realistic systems are often in the “dirty” limit (literally and metaphorically) and the fun ones are typically far from the limits in which any type of perturbative or otherwise approximate analytical theory holds.

  3. Cath@VWXYNot? Says:

    I thought the title said “my thoughts on CPC” at first, and am most disappointed not to find hilarious political ranting.

Leave a Reply

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

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

Google+ photo

You are commenting using your Google+ 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: