Improved Transitional MCMC for Bayesian Inference
MATLAB and Python 3 software for Bayesian inference of engineering models using the improved Transitional Markov Chain Monte Carlo (iTMCMC) method. The TMCMC method samples a sequence of intermediate distributions defined by some tempering parameters that gradually approach the target posterior distribution. Based on the original TMCMC method, the iTMCMC algorithm implements the following modifications: (1) Adjust the sample weights after each MCMC step in order to reduce the average bias of the estimation of the model evidence; (2) Apply a burn-in period in the MCMC step in order to improve the posterior approximation; and (3) Adaptive selection of the scale of the proposal distribution of the MCMC algorithm in order to achieve a near-optimal acceptance rate (this requires performing the Bayesian inference in an underlying standard Gaussian space).
Required input:
- Number of samples per level
- Burn-in period
- Logarithm of the likelihood function passed as a handle to a user-defined MATLAB function
- Prior distribution defined as an object of the ERANataf class
The software returns:
- Samples from the posterior distribution in standard space
- Samples from the posterior distribution in original space
- Intermediate values of the tempering parameter
- An estimate of the model evidence
Requirements
MATLAB, incl. Statistical toolbox, ERADist and ERANataf probability distribution classes
Python 3
Documentation & background
Ching. J, Chen Y.-C. (2007): Transitional Markov chain Monte Carlo method for Bayesian model updating,model class selection, and model averaging. Journal of Engineering Mechanics, ASCE, 133(7): 816-832.
Betz W., Papaioannou I., Straub D. (2016): Transitional Markov chain Monte Carlo: Observations and improvements. Journal of Engineering Mechanics, ASCE, 142(5): 04016016.