Parallel Tempering
Summary
This is a Fortran90 source code for multi-dimensional optimization (derivative free) and probabilistic sampling. Parallel tempering is a technique for accelerating convergence of Markov chains in multi-dimensional sampling. It can be used for Bayesian probabilistic sampling or optimization (misfit function fitting). This software is intended as a wrapper code to be used in conjunction with the users own Markov chain code, or using the internal MCMC algorithm supplied. It can be used for general Bayesian sampling or optimization. Examples of both appear in Sambridge (2013) along with full details of the implementation used here. Examples of driver routines calling the PT library are supplied with the code.
A pdf manual is available here and provides information on installation and usage. A copy of the manual is included in the download package.
Download
The code can be downloaded here. The package includes the above pdf manual and some examples. You will need to register with iEarth prior to download.
The contents of this package is provided under the terms of the GPL licence
Enquires should be directed to the author Malcolm Sambridge.
References
Sambridge, M., 2013. A Parallel Tempering algorithm for probabilistic sampling and multi-modal optimization, Geophys. J. Int., doi: 10.1093/gji/ggt342
Dosso, S. E., Holland, C. W. and Sambridge, M., 2012. Parallel tempering in strongly nonlinear geoacoustic inversion, J. Acoust. Soc. Am., 132, Issue 5, 3030-3040. doi:10.1121/1.4757639