|
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 - #306525 |
|
Title:
|
Bayesian Computational Methods for Models in Geosciences
|
|
Author(s):
|
Alejandro Villagran*+ and Gabriel Huerta
|
|
Companies:
|
University of New Mexico and University of New Mexico
|
|
Address:
|
415 Humanities Building, Albuquerque, NM, 87131-0001,
|
|
Keywords:
|
multiple very fast simulated annealing ; adaptive Metropolis-Hastings ; geophysical inversion problem ; Bayesian inference
|
|
Abstract:
|
In this talk, we will describe the development and application of different intensive computational techniques that arise from Bayesian approaches to geosciences models. These methods include algorithms such as multiple, very fast, simulated annealing; adaptive Metropolis-Hastings (M-H); parallel tempering; and the multiple try method. We compare the performance of each algorithm and show both advantages and drawbacks in the context of estimating an earthquake epicenter location and in climate model parameter estimation. Mainly, we focus on two aspects: optimization of the posterior density and optimal uncertainty estimation.
|
