The traditional group picture at the end of Eric Moulines’ 60th anniversary celebration, at IHP, Paris. Some of the participants had already left (and I am carefully hidding in the background). Among the celebrating talks reflecting the huge thematic diversity of EM’s carreer, Patrick Flandrin gave a great historical account of a certain Édouard-Léon Scott de Martinville and his invention of a sound recording device that did not meet with the same success as the later phonograph by Edison. A song he had registered in 1860 was retrieved in 2008 by a team of the Lawrence Berkeley National Laboratory, making it the earliest known intelligible voice recording in existence! Jean-François Cardoso explained how the team at Institut d’Astrophysique de Paris produced a near optimal estimate of the Cosmic Microwave Background (CMB) by linear projections preserving normality. Sara Filippi exposed a variational Bayes approach to selecting groups of variables in a GLM. Gareth Roberts illustrated retrospective sampling with his recent foray with Jeff Rosenthal in the lack of uniformity in the FIFA World Cup draws. Anatoli Iouditski spoke about a recent work on polyhedral estimation in statistical linear inverse problems. And Elisabeth Gassiat strolled through recent works on inference for hidden Markov models, including one at NeurIPS 2021 with Aapo Hyvärinen and others on nonlinear ICA. This was quite a fun meeting, with plenty of anecdotes and a few older pictures (even though I could not find any prior to 2005, which may have been the year I bought my first digital camera!)
Archive for Eric Moulines
sweet 60’s
Posted in Kids, pictures, Statistics, University life with tags Édouard-Léon Scott de Martinville, celebrations, CMB, cosmic microwave background, Eric Moulines, FIFA, football World Cup, group picture, IHP, Institut Henri Poincaré, Lawrence Berkeley National Laboratory, NeurIPS 2021, NYT, Paris, recording, The New York Times, Thomas Edison, Zeitschrift on October 9, 2023 by xi'anÉric turns 60
Posted in Books, Mountains, Statistics, Travel, University life with tags Aussois, birthday, Dent Parrachée, ERC Synergy Grant, Eric Moulines, Festschrift, French Alps, generative models, IHP, Institut Henri Poincaré, Maurienne, Paris, Vanoise, variational inference on July 14, 2023 by xi'an |
Conférence en l’honneur d’Eric Moulines13-14 sept. 2023 IHP Paris (France)
|
There will be a conference / Festschrift held for the 60th birthday of Eric Moulines [longtime friend, coauthor, ERC Synergy co-PI, whom I almost killed when missing a catch on Dent Parrachée!] at Institut Henri Poincaré, Paris, on 13 and 14 September 2023. Registration is free, but compulsory.
These two days will celebrate the variety of the research topics covered by Eric Moulines over thirty years and the broad and thorough impact he had on many different scientific communities. In particular, through more than 100 papers published in highly selective journals, he has made essential contributions in as many fields as statistical signal processing, time series analysis, inference in partially observed models, non-linear filtering, computational statistics, Markov Chain Monte Carlo, stochastic optimization methods, as well as hot topics related to artificial intelligence like generative models, variational inference or Bayesian machine learning.
MCMC, variational inference, invertible flows… bridging the gap?
Posted in Books, Mountains, Running, Statistics, Travel, University life with tags acronym, alpine climbing, École Polytechnique, biking, Eric Moulines, flat tyre, Hamiltonian Monte Carlo, HMC, invertible flows, MCMC, Pic du Midi d'Ossau, prairie, Pyrennées, teaching, University of Warwick, variational autoencoders, variational inference, webinar on October 2, 2020 by xi'an
Two weeks ago, my friend [see here when climbing Pic du Midi d’Ossau in 2005!] and coauthor Éric Moulines gave a very interesting on-line talk entitled MCMC, Variational Inference, Invertible Flows… Bridging the gap?, which was merging MCMC, variational autoencoders, and variational inference. I paid close attention as I plan to teach an advanced course on acronyms next semester in Warwick. (By acronyms, I mean ABC+GAN+VAE!)
The notion in this work is that variational autoencoders are based on over-simple mean-field variational distributions, that usually produce a poor approximation of the target distribution. Éric and his coauthors propose to introduce a Metropolis step in the VAE. This leads to a more general notion of Markov transitions and a global balance condition. Hamiltonian Monte Carlo can be used as well and it improves the latent distribution approximation, namely the encoder, which is surprising to me. The steps of the Markov kernel produce a manageable transform of the initial mean field approximation, a random version of the original VAE. Manageable provided not too many MCMC steps are implemented. (Now, the flow of slides was much too fast for me to get a proper understanding of the implementation of the method, of the degree of its calibration, and of the computing cost. I need to read the associated papers.)
Once the talk was over, I went back to changing tires and tubes, as two bikes of mine had flat tires, the latest being a spectacular explosion (!) that seemingly went through the tire (although I believe the opposite happened, namely the tire got slashed and induced the tube to blow out very quickly). Blame the numerous bits of broken glass over bike paths.
Markov Chains [not a book review]
Posted in Books, pictures, Statistics, University life with tags book review, concentration inequalities, coupling, Eric Moulines, irreducibility, Markov chain and stochastic stability, Markov chain Monte Carlo, Markov chains, MCMC convergence, probability theory, Randal Douc, Richard Tweedie, Sean Meyn, Wasserstein distance on January 14, 2019 by xi'an
Nonlinear Time Series just appeared
Posted in Books, R, Statistics, University life with tags book review, CHANCE, EM algorithm, Eric Moulines, Markov chains, MCMC, Monte Carlo Statistical Methods, nonlinear time series, particle filters, pMCMC, R, Randal Douc, sequential Monte Carlo, simulation, statistical inference, time series on February 26, 2014 by xi'an
My friends Randal Douc and Éric Moulines just published this new time series book with David Stoffer. (David also wrote Time Series Analysis and its Applications with Robert Shumway a year ago.) The books reflects well on the research of Randal and Éric over the past decade, namely convergence results on Markov chains for validating both inference in nonlinear time series and algorithms applied to those objects. The later includes MCMC, pMCMC, sequential Monte Carlo, particle filters, and the EM algorithm. While I am too close to the authors to write a balanced review for CHANCE (the book is under review by another researcher, before you ask!), I think this is an important book that reflects the state of the art in the rigorous study of those models. Obviously, the mathematical rigour advocated by the authors makes Nonlinear Time Series a rather advanced book (despite the authors’ reassuring statement that “nothing excessively deep is used”) more adequate for PhD students and researchers than starting graduates (and definitely not advised for self-study), but the availability of the R code (on the highly personal page of David Stoffer) comes to balance the mathematical bent of the book in the first and third parts. A great reference book!
