Archive for Luminy
CIRM, Luminy, 1995
Posted in Statistics with tags CIRM, ESF, European Science Foundation, HSSS, Luminy, memories, MRC Biostatistics Unit, Sylvia Richardson, University of Cambridge on May 12, 2023 by xi'anStatMathAppli, Fréjus, France [18-22 September 2023]
Posted in Statistics, Travel, University life with tags CIRM, ENSAE, Luminy, optimal transport, retrospective Monte Carlo, StatMathAppli 2023, University of Warwick on April 13, 2023 by xi'anThe bi-yearly StatMathAppli conference will take place next September in Fréjus, France, with guest lecturers Marco Cuturi (ENSAE & Apple ML Research) who will a lecture on “Optimal Transport: From Theory to Tweaks, Computations and Applications in Machine Learning”, and Gareth O. Roberts (University of Warwick) who will give a lecture on “Topics in Retrospective Simulation”. Registration is open and the number of places is limited. (I got invited to the 2002 edition of this workshop, which took place in CIRM, Luminy.)
Fusion at CIRM
Posted in Mountains, pictures, Statistics, Travel, University life with tags ABC, Bayesian non-parametrics, BNP, boar, Chili, CIRM, cold water swimming, data privacy, fusion, Les Calanques, Luminy, Luminy campus, Méditerranée, MCMC, Parc National des Calanques, particle filter, SMC, Université Aix Marseille, workshop on October 24, 2022 by xi'anToday is the first day of the FUSION workshop Rémi Bardenet and myself organised. Due to schedule clashes, I will alas not be there, since [no alas!] at the BNP conference in Chili. The program and collection of participants is quite exciting and I hope more fusion will result from this meeting. Enjoy! (And beware of boars, cold water, and cliffs!!!)
control variates [seminar]
Posted in pictures, Statistics, Travel, University life with tags Athens, CIRM, control variates, convergence of Gibbs samplers, Langevin MCMC algorithm, London, Luminy, MALA, Marseiile, MCMC, Metropolis adjusted Langevin algorithm, Poisson equation, random walk Metropolis algorithm, UCL, University of Warwick on November 5, 2021 by xi'anToday, Petros Dellaportas (whom I have know since the early days of MCMC, when we met in CIRM) gave a seminar at the Warwick algorithm seminar on control variates for MCMC, reminding me of his 2012 JRSS paper. Based on the Poisson equation and using a second control variate to stabilise the Monte Carlo approximation do the first control variate. The difference with usual control variates is finding a first approximate G(x)-q(y|x)G(Y) to F-πF. And the first Poisson equation is using α(x,y)q(y|x) rather than π. Then the second expands log α(x,y)q(y|x) to achieve a manageable term.
Abstract: We provide a general methodology to construct control variates for any discrete time random walk Metropolis and Metropolis-adjusted Langevin algorithm Markov chains that can achieve, in a post-processing manner and with a negligible additional computational cost, impressive variance reduction when compared to the standard MCMC ergodic averages. Our proposed estimators are based on an approximate solution of the Poisson equation for a multivariate Gaussian target densities of any dimension.
I wonder if there were a neural network version that would first build G from scratch and later optimise it towards solving the Poisson equation. As in this recent arXival I haven’t read (yet).