Online Program Home
My Program

Abstract Details

Activity Number: 253 - Contributed Poster Presentations: Section on Statistical Computing
Type: Contributed
Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistical Computing
Abstract #330726
Title: Continuous Tempering Through Path Sampling
Author(s): Yuling Yao* and Andrew Gelman
Companies: Columbia Univ and Columbia University
Keywords: Bayesian computation; Hamiltonian Monte Carlo; simulated tempering; path sampling; normalizing constant

Hamiltonian Monte Carlo may face challenges when sampling from a multi-modal distribution. We propose a new tempering algorithm by introducing a continuous auxiliary temperature parameter and conduct joint HMC jumps in the joint space. The path sampling gives a smoothed estimation of normalizing constant, while the adaptive procedure makes more reliable estimation for the partition function of the entire path. We compare the results with other tempering methods and find the proposed approach to be optimal in simulation examples.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program