|
Activity Number:
|
236
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #306469 |
|
Title:
|
Sampling Importance Resampling for Computer Model Inverse Problems
|
|
Author(s):
|
Matt Taddy*+ and Bruno Sanso and Herbert Lee
|
|
Companies:
|
University of California, Santa Cruz and University of California, Santa Cruz and University of California, Santa Cruz
|
|
Address:
|
231 Blackburn Street, Santa Cruz, CA, 95060,
|
|
Keywords:
|
inverse problems ; computer models ; sampling importance resampling
|
|
Abstract:
|
The classic approach to statistical inverse problems is to develop a surrogate model for the expensive computer simulator and to use this to find the likelihood for sampling from the posterior distribution of inputs. However, in many cases, there is a large bank of simulated values and the need for a faster algorithm that does not require predicted output at new locations. We develop a sampling importance resampling (SIR) algorithm that works in conjunction with kernel density estimation to resample from the original computer output according to the posterior distribution of input values. We will examine and compare the performance of our algorithm in examples that include multivariate output data from a nondeterministic climate model simulator.
|
