UNIGE document Scientific Article
previous document  unige:96975  next document
add to browser collection

Fast Langevin based algorithm for MCMC in high dimensions

Durmus, Alain
Roberts, Gareth O.
Zygalakis, Konstantinos C.
Published in Annals of Applied Probability. 2017, vol. 27, no. 4, p. 2195-2237
Abstract We introduce new Gaussian proposals to improve the efficiency of the standard Hastings-Metropolis algorithm in Markov chain Monte Carlo (MCMC) methods, used for the sampling from a target distribution in large dimension $d$. The improved complexity is $mathcal{O}(d^{1/5})$ compared to the complexity $mathcal{O}(d^{1/3})$ of the standard approach. We prove an asymptotic diffusion limit theorem and show that the relative efficiency of the algorithm can be characterised by its overall acceptance rate (with asymptotical value 0.704), independently of the target distribution. Numerical experiments confirm our theoretical findings.
arXiv: 1507.02166
Full text
Research group Analyse numérique
(ISO format)
DURMUS, Alain et al. Fast Langevin based algorithm for MCMC in high dimensions. In: Annals of Applied Probability, 2017, vol. 27, n° 4, p. 2195-2237. https://archive-ouverte.unige.ch/unige:96975

142 hits



Deposited on : 2017-09-25

Export document
Format :
Citation style :