Archive for pseudo-data

ABC à Montréal

Posted in Kids, pictures, Running, Statistics, Travel, University life with tags , , , , , , , , , , , , , , , , on December 13, 2014 by xi'an

Montreal1So today was the NIPS 2014 workshop, “ABC in Montréal“, which started with a fantastic talk by Juliane Liepe on some exciting applications of ABC to the migration of immune cells, with the analysis of movies involving those cells acting to heal a damaged fly wing and a cut fish tail. Quite amazing videos, really. (With the great entry line of ‘We have all cut  a finger at some point in our lives’!) The statistical model behind those movies was a random walk on a grid, with different drift and bias features that served as model characteristics. Frank Wood managed to deliver his talk despite a severe case of food poisoning, with a great illustration of probabilistic programming that made me understand (at last!) the very idea of probabilistic programming. And  Vikash Mansinghka presented some applications in image analysis. Those two talks led me to realise why probabilistic programming was so close to ABC, with a programming touch! Hence why I was invited to talk today! Then Dennis Prangle exposed his latest version of lazy ABC, that I have already commented on the ‘Og, somewhat connected with our delayed acceptance algorithm, to the point that maybe something common can stem out of the two notions. Michael Blum ended the day with provocative answers to the provocative question of Ted Meeds as to whether or not machine learning needed ABC (Ans. No!) and whether or not machine learning could help ABC (Ans. ???). With an happily mix-up between mechanistic and phenomenological models that helped generating discussion from the floor.

The posters were also of much interest, with calibration as a distance measure by Michael Guttman, in continuation of the poster he gave at MCMski, Aaron Smith presenting his work with Luke Bornn, Natesh Pillai and Dawn Woodard, on why a single pseudo-sample is enough for ABC efficiency. This gave me the opportunity to discuss with him the apparent contradiction with the result of Kryz Łatunsziński and Anthony Lee about the geometric convergence of ABC-MCMC only attained with a random number of pseudo-samples… And to wonder if there is a geometric versus binomial dilemma in this setting, Namely, whether or not simulating pseudo-samples until one is accepted would be more efficient than just running one and discarding it in case it is too far. So, although the audience was not that large (when compared with the other “ABC in…” and when considering the 2500+ attendees at NIPS over the week!), it was a great day where I learned a lot, did not have a doze during talks (!), [and even had an epiphany of sorts at the treadmill when I realised I just had to take longer steps to reach 16km/h without hyperventilating!] So thanks to my fellow organisers, Neil D Lawrence, Ted Meeds, Max Welling, and Richard Wilkinson for setting the program of that day! And, by the way, where’s the next “ABC in…”?! (Finland, maybe?)

a pseudo-marginal perspective on the ABC algorithm

Posted in Mountains, pictures, Statistics, University life with tags , , , , , , , , on May 5, 2014 by xi'an

ridge6

My friends Luke Bornn, Natesh Pillai and Dawn Woodard just arXived along with Aaron Smith a short note on the convergence properties of ABC. When compared with acceptance-rejection or regular MCMC. Unsurprisingly, ABC does worse in both cases. What is central to this note is that ABC can be (re)interpreted as a pseudo-marginal method where the data comparison step acts like an unbiased estimator of the true ABC target (not of the original ABC target, mind!). From there, it is mostly an application of Christophe Andrieu’s and Matti Vihola’s results in this setup. The authors also argue that using a single pseudo-data simulation per parameter value is the optimal strategy (as compared with using several), when considering asymptotic variance. This makes sense in terms of simulating in a larger dimensional space but what of the cost of producing those pseudo-datasets against the cost of producing a new parameter? There are a few (rare) cases where the datasets are much cheaper to produce.

JSM 2011 [reflections]

Posted in pictures, Statistics, Travel, University life with tags , , , , , , on August 5, 2011 by xi'an

The meeting is now over and I should be busy packing rather than writing this post. This has been a highly busy week, with many meetings on the side, while working at night by refraining from fighting jetlag (as usual), so I should also let things rest rather than letting a sort of post-meeting melancholia express itself… As after last year JSM… Anyway, here are some of my raw reflections on JSM 2011.

On the positive side, I attended many exciting sessions, either because they were bringing new perspectives to me—maybe the keyword I will carry back from Miami Beach is pseudo-data— or because they exhibited a comprehensive and influential perspective on a domain (I am mostly thinking of David Cox’s and IMS medallion lectures). I met new people (including the editorial board of CHANCE!) and old friends (the Bayesian mixer was too short!),  delivered the rewards for the Mitchell Prize to a great paper on galaxy formation by Ian Vernon, Michael Goldstein, and Richard Bower, had several conclusive “business” meetings (and a few disappointing ones to keep the balance right!). I even managed to stick a working session into the tight program (although I wish it had been at another time in the day as I was partly dozing away…) I also enjoyed a terrific Cuban dinner in Versailles (!) and managed to take a few satisfactory pictures of sunrise (to be imposed on the readers in the coming days, I afraid!).

On the down side, I attended too many sessions with a very small audience, although the talks deserved better. Maybe due to the humongous size of the convention center, maybe due to the lesser attendance, maybe due to the strong attraction of the nearby beaches, I generally had a feeling of being in a small meeting. As noted by Julien, having so many parallel sessions is both an organisational nightmare and an academic absurdity. Besides forcing attendees to make choices between sessions (the worst case being the Savage award delivered during my Bayesian model assessment session!), it dulls the attractiveness of the meeting and the relevance of the talks. It is certainly not going to happen, but JSM should have a stronger filter for proposed talks in order to avoid contributed sessions where the only attendees are the five speakers plus the chair! It should also do something about the last day sessions: since canceling the last day of the conference is not possible (if only because there would be another last day!), inventing an attractive programme for the last sessions would anchor more attendees till the end. A national (and international) meeting of this size is an enormously expensive monster, in terms of costs both to the universities and companies (especially in Miami Beach!), and to the environment. The RSS went the major step of canceling the yearly meeting this year and, although the size of the meeting is not the same, the statistical societies involved in JSM could maybe consider alternatives. One way could be to encourage videotransmission of talks (of course, this would not reduce the number of talks, but impact the size of the audience. I tried to give a talk at MCQMC next year this way, as flying to Sydney for three days did not sound realistic, but this proposal was not received positively!) There is no obvious solution to this issue, otherwise it would have been found, but this feeling of somehow wasting enormous amounts of money in an uncertain economy contributes to my melancholia….

On a more personal [down]side, having to watch for Emily and planning for alternative vacation plans did not help with my stress level! (At this stage the hurricane warning is off. And so are we.) The constant heat and humidity did not either, even though I knew in advance it would be a problem and decided not to whine about it (at least on this blog…)  The cost of living in Miami Beach however came as a surprise, although it may explain for the lower attendance this year. (Having rented an apartment across the street from the convention center was a partial solution to both problems, though.)