Choose France – CNRS AI Rising Talents Programme for exceptionally talented early-career AI researchers

The French national scientific research institute, CNRS, is launching a one-shot call for hiring early-career  AI researchers on a 5 year 100% research position in any of the CNRS research units in France. Including ours in Paris Dauphine. The offer is quite generous in that the attached grant provides a full professor salary and support to hire postdocs and PhD students. Early career means two to seven years after PhD. The deadline is 30 November.

pr[AI]rie day

convergences of MCMC and unbiasedness

During his talk on unbiased MCMC in Dauphine today, Pierre Jacob provided a nice illustration of the convergence modes of MCMC algorithms. With the stationary target achieved after 100 Metropolis iterations, while the mean of the target taking much more iterations to be approximated by the empirical average. Plus a nice connection between coupling time and convergence. Convergence to the target.During Pierre’s talk, some simple questions came to mind, from developing an “impatient user version”, as in perfect sampling, in order  to stop chains that run “forever”,  to optimising parallelisation in order to avoid problems of asynchronicity. While the complexity of coupling increases with dimension and the coupling probability goes down, the average coupling time varies but an unexpected figure is that the expected cost per iteration is of 2 simulations, irrespective of the chosen kernels. Pierre also made a connection with optimal transport coupling and stressed that the maximal coupling was for the proposal and not for the target.

applied Bayesian statistical modelling (PhD course at CREST)

Next month, Kerrie Mengersen (QUT, Brisbane, Australia, and visiting us at CREST and Paris-Dauphine this coming May) will give a PhD course at CREST on the theme of applied Bayesian statistical modelling.

Here is her abstract:

Bayesian hierarchical models are now widely used in addressing a rich variety of real-world problems. In this course, we will examine some common models and the associated computational methods used to solve these problems, with a focus on environmental and health applications.

Two types of hierarchical models will be considered, namely mixture models and spatial models. Computational methods will cover Markov chain Monte Carlo, Variational Bayes and Approximate Bayesian Computation.

Participants will have the opportunity to implement these approaches using a number of datasets taken from real case studies, including the analysis of digital images from animals and satellites, and disease mapping for medicine and biosecurity.

The classes will take place at ENSAE, Paris, on May 3, 10 (14:00, Amphi 2), 14, and 21 (11:00, Room S8). (The course is open to everyone and free of charge, but registrations are requested, please contact Nadine Guedj.)

From my office window