after-dinner at Trinity [jatp]

at the Isaac Newton Institute [talks]

Here are the slides I edited this week [from previous talks by Pierre and Epstein] for the INI Workshop on scalable inference, in connection with our recently completed and submitted paper on ABC with Wasserstein distances:

at the Isaac Newton Institute

Le Monde on the “dangers” of mathematics

“La responsabilité des mathématiciens semble engagée.”

This post is presumably aiming at a very small (French speaking) audience, but Le Monde published a central Science leaflet this week on the dangers of using uncontrolled mathematical modelling. Resulting in a mismatch of platitudes and absurdities. Blaming mathematicians for about every misappropriate use of mathematics and even more statistics, from the lack of reproducibility in published psychology studies and the poor predictions of flu epidemics by Google to the sub-prime crisis and the prosecutor fallacy. Quoting judicial miscarriages like the case of Lucy de Berk when the statistical arguments were administrated by a psychologist, while a statistician, Richard Gill, was instrumental in reopening the case by demonstrating those arguments were wrong. Objecting to the use of logistic regression for profiling inmates on the probability of recidivism. &tc., &tc… The only item of interest in this really poor article is the announcement of a semester workshop at the Isaac Newton Institute on the use of mathematics in criminal sciences. Which after verification is a workshop on probability and statistics in forensic sciences. With Richard Gill as one of the organisers.

an all-inclusive Cambridge experience

My trip to Cambridge last week was very enjoyable: I have visited Cambridge and the (terrific) Isaac Newton institute enough times in the past twenty years to know my way around and to look forward my next visit… However, having the opportunity of staying in St John’s College with a room next to the Cam river and attending a formal dinner in the magnificent Dinning Hall was a wonderful new experience!

I am always amazed (and obviously envious) of the facilities offered to both students and faculty living in the colleges (not mentioning of course the rich intellectual atmosphere both in colleges and at the faculty of mathematics, expressed e.g. by an astounding list of seminars). My seminar on ABC model choice was well-attended and discussed, despite being at 4pm a Friday afternoon (maybe the beer social in the faculty upstairs right after helped!). And I had time to work quietly on multiple-try MCMC, following thoughts raised both by the comments on the related post and the GPU meeting in Warwick. Add great weather and the opportunity to go for a run next to the rowers in the morning to make for a great (short) trip.

2012, Turing year

Buying the special issue of La Recherche on “La révolution des mathématiques”, I discovered that this is the Alan Turing Year in celebration of the 100th anniversary of Turing‘s birth. The math department at the University of Leeds has a webpage on all the events connected with this celebration. From all over the World. (There is even a Turing relay in Cambridge, unfortunately it is not open to the general public… Unless you are attending the Isaac Newton Institute at the time.) Quite fitting a tribute. (Given Turing’s contributions to Bayesian analysis, as depicted e.g. in the theory that would not die, ISBA could have included a special session in the ISBA 2012 meeting in Kyoto. I will certainly dedicate the session I co-organise there on parallel computing to his memory.)

Rare events in Cambridge

This morning, I got a mass email about an interesting workshop at the Isaac Newton Institute in Cambridge about rare event simulation:

RESIM 2010 is the 8th workshop in a series of successful events held on the same topic in Aachen (Germany) 1997, Enschede (The Netherlands) 1999, Pisa (Italy) 2000, Madrid (Spain) 2002, Budapest (Hungary) 2004, Bamberg (Germany) 2006, and Rennes (France) 2008.

It covers all aspects of rare event simulation ranging from purely theoretical developments to practical applications. The objective is to provide a forum for researchers and practitioners working in different locations and on different applications to present recent results, exchange ideas, and discuss open problems and new direction. While contributed talks are encouraged, one need not present a paper in order to participate in this event.

The 2010 meeting is being organized as part of a broader simulation workshop sponsored by the Newton Institute at Cambridge University. The RESIM event will dominate the first two days of the workshop (Monday, June 21 and Tuesday, June 22), while the third day (Wednesday, June 23) will consider simulation topics lying outside the rare-event domain. RESIM 2010 participants will be automatically registered in this broader three day event.

If you have an interest in presenting a paper at RESIM 2010, please note that the conference has a particular interest in soliciting papers that present advances, new results and applications in the field of rare-event simulation (especially those that make a contribution to some area of network models and/or the communication sciences, given the overall theme of the Newton Institute programme of which this is a part). The scope of the workshop includes (but is not limited to):

* Importance sampling based simulation techniques
* Rare event simulation techniques based on splitting sample paths (e.g., the RESTART method)
* Other novel approaches to rare event simulation
* Rare event simulation of heavy-tailed and long range dependent processes
* Large deviations theory

The Program Committee for RESIM 2010 consists of: Søren Asmussen (Aarhus U, Denmark), José Blanchet (Columbia U, USA), Sergey Foss (Herriott-Watt U, UK), Peter Glynn (Stanford U, USA), Victor Nicola (Twente U, Netherlands), and Bruno Tuffin (INRIA, France). The expectation is that at least one co-author of each accepted paper will participate in the RESIM conference.

I wish I could go but with three other meetings planned for June, this is getting a bit too tight!