de-MCM’d

This morning I received a message from the MCM 23 conference organisers that my registration [submitted two months ago] was declined for lack of room! I wonder why the organisers did not opt for broadcasting in a second amphitheater, as was done for ISBA in Edinburgh.

Unfortunately, we have attained the maximal capacity of the amphitheater where the plenary talks will take place (this is the largest amphitheater that one can rent on the Jussieu campus). This amphitheater capacity was significantly larger than the number of attendees of previous MCM conferences. We feel really sorry that we can’t confirm your registration to MCM2023.

Which is pretty frustrating given that the program is of the highest standards and that many friends, coauthors, students, of mine’s are giving talks there. Being a local I’ll try to gatecrash some talks, of course, but I would not bet on my chances, unless I can borrow a badge!

on control variates

A few months ago, I had to write a thesis evaluation of Rémi Leluc’s PhD, which contained several novel Monte Carlo proposals on control variates and importance techniques. For instance, Leluc et al. (Statistics and Computing, 2021) revisits the concept of control variables by adding a perspective of control variable selection using LASSO. This prior selection is relevant since control variables are not necessarily informative about the objective function being integrated and my experience is that the more variables the less reliable the improvement. The remarkable feature of the results is in obtaining explicit and non-asymptotic bounds.

The author obtains a concentration inequality on the error resulting from the use of control variables, under strict assumptions on the variables. The associated numerical experiment illustrates the difficulties of practically implementing these principles due to the number of parameters to calibrate. I found the example of a capture-recapture experiment on ducks (European Dipper) particularly interesting, not only because we had used it in our book but also because it highlights the dependence of estimates on the dominant measure.

Based on a NeurIPS 2022 poster presentation Chapter 3 is devoted to the use of control variables in sequential Monte Carlo, where a sequence of importance functions is constructed based on previous iterations to improve the approximation of the target distribution. Under relatively strong assumptions of importance functions dominating the target distribution (which could generally be achieved by using an increasing fraction of the data in a partial posterior distribution), of sub-Gaussian tails of an intractable distribution’s residual, a concentration inequality is established for the adaptive control variable estimator.

This chapter uses a different family of control variables, based on a Stein operator introduced in Mira et al. (2016). In the case where the target is a mixture in IR^d, one of our benchmarks in Cappé et al. (2008), remarkable gains are obtained for relatively high dimensions. While the computational demands of these improvements are not mentioned, the comparison with an MCMC approach (NUTS) based on the same number of particles demonstrates a clear improvement in Bayesian estimation.

Chapter 4 corresponds to a very recent arXival and presents a very original approach to control variate correction by reproducing the interest rate law through an approximation using the closest neighbor (leave-one-out) method. It requires neither control function nor necessarily additional simulation, except for the evaluation of the integral, which is rather remarkable, forming a kind of parallel with the bootstrap. (Any other approximation of the distribution would also be acceptable if available at the same computational cost.) The thesis aims to establish the convergence of the method when integration is performed by a Voronoi tessellation, which leads to an optimal rate of order n^-1-2/d for quadratic error (under conditions of integrand regularity). In the alternative where the integral must be evaluated by Monte Carlo, this optimality disappears, unless a massive amount of simulations are used. Numerical illustrations cover SDEs and a Bayesian hierarchical modeling already used in Oates et al. (2017), with massive gain in both cases.

Research fellow chAIrs in Grenoble [reposted]

MIAI, the Grenoble Multidisciplinary Institute in Artificial Intelligence , is opening three research fellow chairs in AI reserved to persons who have spent most of their research career outside France. To be eligible, candidates must hold a PhD from a non-French university obtained after January 2014 for male applicants and after 2014-n, where n is the number of children, for female applicants. They must also have spent more than two thirds of their research career since the beginning of their PhD outside France. These research fellow chairs aim to to address important and ambitious research problems in AI-related fields and will partly pave the way for the future research to be conducted in MIAI. Successful candidates will be appointed by MIAI and will be allocated, for the whole duration of the chair, a budget of 250k€ covering PhD and/or postdoc salaries, internships, travels. The deadline for applications is 11 March 2023, see here for details.

another round of mostly useless road death statistics [and a terrible graph]

Another terrible report on (French) road accidents and deaths Le Monde pointed to. The entire analysis does not consider once the number of people on the roads or the death per kilometer ratio. Which makes the absolute figures as those represented in this ugly graph hard to comment. For instance, the number of persons cycling to work has increased more than the number of bike deaths. (And, contrary to a urban myth, cycling in Paris should not be considered as a extreme sport: only one  [too many] cyclist died there in 2022.) I also find surprising the (a)symmetry in the age distributions of (overall) road deaths,

since the percentages of evolution between 2019 and 2022 almost exactly compensate for one to the next across the age groups. Any significance in these figures? The statistics that makes the most sense in the report is the comparison of counties where the 90km/h speed limit was reinstated and those where it stayed at 80km/h: an increase of 1% versus a decrease of 2%… As signaled by Le Monde car doors are bike killers: when getting off a car, use your right hand to open the driver’s door (except in Australia, Britain, Japan and 72 other left-hand driving countries!!).

MCM in Paris, 2023

The next MCM conference takes place in (downtown) Paris next 26-30 June. Deadlines are 31 December for mini-symposia/invited sessions and 28 February for contributed talks/posters. I appreciate very much the effort in lowering the registration fees to 80€ for students and 170€ for others, whilst including lunches into the deal! (The view of Paris in the above logo is actually taking from Paris Jussieu campus.)