|
Activity Number:
|
104
- Advances in Bayesian Analysis of Computer Models
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 8, 2022 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #323537
|
|
|
Title:
|
Theory of Variational Bayes Computer Models
|
|
Author(s):
|
Shrijita Bhattacharya* and Mookyong Son and Vojtech Kejzlar and Tapabrata Maiti
|
|
Companies:
|
Michigan State University and Michigan State University and Skidmore College and Michigan State University
|
|
Keywords:
|
Variational Inference;
Computer Models;
Posterior Contraction;
Calibration
|
|
Abstract:
|
Computer models are widely used for modeling mathematical experiments. For Bayesian calibration, a Gaussian process emulator and a Gaussian process model for the systematic discrepancy between the computer simulation and the underlying physical process is used. In this work, we propose a variational Bayes algorithm as a scalable alternative to the traditional computationally intensive Markov Chain Monte Carlo (MCMC) approach to Bayesian calibrated computer models. We also provide rigorous theoretical justification of the proposed approach by stablishing posterior contraction rates of the estimated physical process. The numerical performance is presented in context of simulation studies and a real data example.
|
Authors who are presenting talks have a * after their name.
