Research Title
Research Website
www.nanoHUB.org
Keywords
online simulation, nanoHUB, bayesian calibration, curve fitting, uncertainty quantification
Presentation Type
Event
Research Abstract
Fitting a model to data is common practice in many fields of science. The models may contain unknown parameters and often, the goal is to obtain good estimates of them. A variety of methods have been developed for this purpose. They often differ in complexity, efficiency and accuracy and some may have very limited applications. Bayesian inference methods have recently become popular for the purpose of calibrating model's parameters. The way they treat unknown quantities is completely different from any classical methods. Even though the unknown quantity is a constant, it is treated as a random variable and the desired outcome is it's probability distribution. Good estimates and confidence intervals can then be easily produced from probability distributions. Anohter important feature of Bayesian inference is the ability to include prior knowledge in the calculations. However, Bayesian inference has to be done computationally as it involves solving multidimensional integrals. The Bayesian Calibration tool is an easy-to-use, well documented tool to efficiently carry out the calculations of the calibration process. The tool is open-source and uses fast Markov Chain Monte Carlo (MCMC) algorithms. The tool is run on nanoHUB, making it easily accessible without installing any software, etc. Given data and a model, the tool performs MCMC simulation of the model and returns the Bayesian posterior probability distributions of the model's unknown parameters.
Session Track
Nanotechnology
Recommended Citation
Sveinn Palsson, Martin Hunt, and Alejandro Strachan,
"Bayesian Calibration Tool"
(August 7, 2014).
The Summer Undergraduate Research Fellowship (SURF) Symposium.
Paper 19.
https://docs.lib.purdue.edu/surf/2014/presentations/19
Included in
Nanoscience and Nanotechnology Commons, Numerical Analysis and Scientific Computing Commons
Bayesian Calibration Tool
Fitting a model to data is common practice in many fields of science. The models may contain unknown parameters and often, the goal is to obtain good estimates of them. A variety of methods have been developed for this purpose. They often differ in complexity, efficiency and accuracy and some may have very limited applications. Bayesian inference methods have recently become popular for the purpose of calibrating model's parameters. The way they treat unknown quantities is completely different from any classical methods. Even though the unknown quantity is a constant, it is treated as a random variable and the desired outcome is it's probability distribution. Good estimates and confidence intervals can then be easily produced from probability distributions. Anohter important feature of Bayesian inference is the ability to include prior knowledge in the calculations. However, Bayesian inference has to be done computationally as it involves solving multidimensional integrals. The Bayesian Calibration tool is an easy-to-use, well documented tool to efficiently carry out the calculations of the calibration process. The tool is open-source and uses fast Markov Chain Monte Carlo (MCMC) algorithms. The tool is run on nanoHUB, making it easily accessible without installing any software, etc. Given data and a model, the tool performs MCMC simulation of the model and returns the Bayesian posterior probability distributions of the model's unknown parameters.
https://docs.lib.purdue.edu/surf/2014/presentations/19