A0352
Title: Performance of the annealed MALA in transience and in stationarity
Authors: Mylene Bedard - University of Montreal (Canada) [presenting]
Abstract: Statistical models have been increasing both in terms of complexity and dimensionality. These models cannot be treated analytically; Markov chain Monte Carlo (MCMC) methods have thus become a device of choice to sample from such target distributions. A generalized version of the Metropolis-adjusted Langevin algorithm (MALA) is introduced that features two tuning parameters: the usual step size and an interpolation parameter that accommodates the dimension of the target distribution. The efficiency of this sampler is theoretically studied by making use of the local- and global-balance concepts of a prior study, and user-friendly tuning guidelines are provided. Although the traditional MALA is theoretically optimal in infinite-dimensional settings, in practice, the annealed MALA remains superior in all contexts. It offers significant efficiency gains both in transience and in stationarity at no extra computational cost. Simulation studies corroborate the findings. In particular, the efficiency of the annealed MALA compares favorably to that of competing algorithms in various Bayesian logistic regression contexts.
