This was a fairly full day at the Structure and uncertainty modelling, inference and computation in complex stochastic systems workshop! After a good one hour run around the Clifton Down, the morning was organised around likelihood-free methods, mostly ABC, plus Arnaud Doucet’s study of methods based on unbiased estimators of the likelihood (à la Beaumont, with the novelty of assessing the inefficiency due to the estimation, really fascinating..). The afternoon was dedicated to graphical models. Nicolas Chopin gave an updated version of his Kyoto talk on EP-ABC where he resorted to composite likelihoods for hidden Markov models, (I then wondered about the parameterisation and the tolerance determination for this algorithm.) Oliver Ratman presented some of the work he did on the flu while in Duke, then move to a new approach for ABC tolerance based on various kinds of testing (which I found clearer than in Kyoto, maybe because I was not jet-lagged!) And I gave my talk on ABC-EL.I found the afternoon session harder to follow, mostly because I always have trouble understanding the motivations and the notations used on these models, albeit fascinating. I remained intrigued by the bidirectional dependence arrow in those graphs for the whole afternoon (even though I think I get it now!) After looking at the few posters presented this afternoon, I went for another short run in Leigh Woods, before joining a group of friends for an Indian dinner at the Brunel Raj. A very full day…!
Archive for Kyoto
I am back in Bristol just a few months after an earlier SuSTain workshop. (During my Spring UK trip to Bristol, Glasgow, and Edinburgh…) The theme of the workshop is Structure and uncertainty modelling, inference and computation in complex stochastic systems. And it enjoys a very rich program over the four days! I am talking about ABC and empirical likelihood, with the following slides I just completed:
Unsurprisingly, those slides borrow both from my earlier talks in Kyoto and Australia, and from Pierre Purdlo’s earlier talk on this paper… (I also added pictures of some of the hikes and climbs Peter Green and I survived together!)
I alas arrived too late for today’s sessions, having to give the opening lecture at my Statistics Master in Paris-Dauphine. (I will also alas miss half of Thursday’s talks!) As I am staying at the Avon Gorge Hotel, just next to the bridge, I took the opportunity of some remaining daylight to go running across Brunel’s bridge and into the nearby park of Leigh Woods. It happened to be very muddy thanks to the torrential rains of the morning, but it was a good way to test my recovering knee (after a minor bike fall last week!) on a long run… And it apparently held, although tomorrow morning run will tell for sure.
Surrounding the great and exciting gathering of Bayesian statisticians in Kyoto last June, several ISBA sections have appeared in the past weeks, as already mentioned on the ‘Og. Along with Anto Mira and Nicolas Chopin (who did most of the organisational work while I was wandering down under!), we discussed about a Bayesian computation section and, thanks to the massive support of the community, we engaged into setting this new section of ISBA, with the help of Peter Green and Håvard Rue. The structure has now been granted an approval stamp by the ISBA highest powers and so here we are with a brand new ISBA Section on Bayesian Computation!!! (A notion I remember discussing with Peter Müller in Valparaiso in…2004!) I think I was the first member to join the section, following the announcement of its official creation by Merlise Clyde… Here is a draft of the call to potential members (along with my own comments):
Over the past twenty years, Bayesian computation has been a tremendous catalyst in Bayesian ideas reaching practitioners – statisticians and non-statisticians alike. It has also providied a fantastic arena for original research in algorithmic statistics and numerical probability, not to mention other fields at the interface. At this more mature stage of its development, at a time where ambitions of statisticians and the expectations on statistics grow, Bayesian computation must remain a major area of research and innovation. Then principled methods of statistical analysis can continue to be both readily available and customarily implemented, as we deal with data on a (much) larger scale, in higher dimensions and with more complex structure.
We invite all ISBA members with (any degree of) interest in computation for Bayesian inference to join the newly created ISBA Section on Bayesian Computation (BayesComp) – and that means both researchers involved in developing new computational methods and associated theory, and users of Bayesian statistical methods interested in implementing, sharing, disseminating, or learning best practice. The purposes of the Section are as multifaceted as the aspects of Bayesian computation, including promoting original research into computational methods for Bayesian inference and decision making, encouraging the use of frontier computational tools among practitioners, the development of adapted software, languages, platforms, and dedicated machines, and translating and disseminating among statisticians methods developed in other disciplines.
To address these purposes, the Section will among other activities organise specific conferences (such as the upcoming MCMSki IV in January 2014), workshops, short courses, webinars, and sessions in other meetings like ISBA and JSM, and will develop and maintain a website of information, tools, and advice as an authoritative central resource for Bayesian computation. The first such resource is already posted: the ISBA Foundation Lecture by Christian Robert on “Approximate Bayesian computation (ABC): Advances and Questions.”[I am definitely not responsible for this inclusion!]
Section dues are only $5 a year or $75 for a Lifetime membership. As part of the Fall (Autumn) Membership Promotion, all new annual memberships will be extended until 31 December, 2013! The section will be holding elections in November, so please join today so that you may participate in choosing the first set of elected officers–and please contact us if you are interested in any of the elected positions! More details to come on the BayesComp section website and the ISBA Bulletin. [You should definitely opt for the Life membership as I did, given that the transaction costs are paid only once! This also means more money for the Section to support younger members towards travel to conferences...]
Welcome to Year 1 BC (BayesComp)!
Nicolas Chopin, Peter Green, Antonietta Mira, Christian Robert and Håvard Rue.
- promote and provide a forum for early career Bayesian researchers.
- organise conferences, workshops, and sessions in other meetings.
- provide social networking tools for early career Bayesian researchers in order to discuss research, exchange ideas with each other, and connect with the Bayesian community at large.
Annual membership dues are $5 and if you join now during ISBA’s Fall Member Promotion, your membership will be extended until December 31, 2013! Membership in the section is limited to current students and junior researchers within five years of completing a degree.
The initial officers of the section are:
- Section chair: Andrew Cron (Term: July 2012-December 2014)
- Chair-elect: to be elected (Term: January 2014-December 2014)
- Program chair: Francesca Ieva (Term: July 2012-December 2013)
- Secretary: Marian Farah (Term: July 2012-December 2013)
- Treasurer: Perla Reyes (Term: July 2012-December 2014)
The first election will be organized concurrently with the 2014 ISBA elections in the Fall of 2013, and will elect the Chair-elect, Program chair, and Secretary who will take office Jan 1, 2014.
Exciting (Bayesian) times!!! Another by-product of the fantastic Kyoto meeting?!
After a short two-day visit to Canberra, I am now heading to Brisbane, last stop of my Australian Tour… I gave my Kyoto talk at ANU, failing to convince my guest Alan Welsh that ABC was a completely justified Bayesian inference method (with the sound argument that it failed the conditionality principle), had a great run in the morning to the top of Black Mountain, witnessing a very nice inversion shown below,
ate a great Indian dinner, and was driven by Alan the next morning to Goulburn for a joint meeting (“Goulburn 19″) between the statisticians of Canberra and of the University of Wollongong. This was a fairly interesting meeting, taking place in a motel conference room. Luxury, luxury!, the room even had a log fire (with the foreseeable consequence of inducing one of my customary and embarrassing dozes!) Actually, this seemed to be a signature of the place as the terrific Trapper’s Bakery across the street also had a fireplace… This meeting was very informal, friendly, and interactive, making me regret I had given there my most technical talk on Rao-Blackwellisation instead of one on ABC model choice.