Adaptive MCMC for infinite dimensional Bayesian inferences |
Date: 2015/10/30 Browse: 671 |
Speaker: Jinglai Li
Time: Oct 30, 10:30am - 11:30am
Location: Room 310, Teaching Center, Zhangjiang Campus
Abstract:
Many scientific and engineering problems require to perform Bayesian inferences in function spaces, in which the unknowns are of infinite dimension. In such problems, most standard Markov Chain Monte Carlo (MCMC) algorithms become arbitrary slow under the mesh refinement, which is referred to as being dimension dependent. On the other hand, in finite dimensional setting, adaptation is often used to improve the sampling efficiency of MCMC iterations. In this talk we present two adaptive MCMC algorithms for the infinite dimensional problems. We show that the proposed algorithms are dimension independent both theoretically and with numerical examples.
Bio:
Jinglai Li (李敬来)
Institute of Natural Sciences Shanghai Jiaotong University
2012 : Distinguished Research Fellow, Shanghai Jiao Tong University
2010 - 2012: Research Associate, MIT
2007 - 2010: Research Associate, Northwestern University
Ph.D.,2007,The State University of New York at Buffalo Research Interests: Scientific Computing, Computational Statistics, Uncertainty Quantification SIST-Seminar 15046
|