Since October 2015, I am a postdoctoral research fellow at School of Mathematics and Statistics & Insight Centre for Data Analytics. I work within the Machine Learning & Statistics group under the supervision of Nial Friel. I am particularly interested in designing adaptive Markov Chain Monte Carlo (MCMC) and noisy Hamiltonian Monte Carlo (HMC) algorithms for Bayesian inference on discrete Markov random fields, a class of intractable statistical models.
Between 2012 and 2015, I was a Ph.D. student at the Institut Montpelliérain Alexander Grothendieck (IMAG, UMR CNRS 5149), the Université de Montpellier department of Mathematics research. I was working within the probability and statistics team under the supervision of Jean-Michel Marin (supervisor), Pierre Pudlo (co-supervisor) and Lionel Cucala (co-supervisor).
During my thesis, I was interested in developing efficient methodologies for (hidden) Markov random fields, especially Ising and Potts models that can be used in many modeling area including image segmentation or genetic analysis. My interest was on the issues of model selection and parameter inference. My thesis disseration is available on http://julien-stoehr.pagesperso-orange.fr.