Gaussian Process-Based Sensitivity Analysis and Bayesian Model Calibration with GPMSA
The Gaussian Process Models for Simulation Analysis (GPMSA) package is a set of functions written in the Matlab programming language aimed at emulating a computer model of a system being studied, calibrating this computer model to observations of the system, and giving predictions of the expected system response. Collectively, these capabilities comprise uncertainty quantification (UQ) in model-supported inference.
This chapter will first discuss some background and motivation for the GPMSA code, then demonstrate the code’s function interfaces in the context of a series of illustrative example problems.
KeywordsBayesian analysis Design and analysis of computer experiments Gaussian process Markov chain Monte Carlo Statistical analysis of computer models
- 1.Graves, T.L.: Automatic step size selection in random walk metropolis algorithms. arXiv preprint, arXiv:11035986 (2011)Google Scholar