View on GitHub

libspatialSEIR

A C++ and OpenCL framework for fast Bayesian spatio-temporal compartmental epidemic modeling.

Disclaimer

This software is the focus of my dissertation work, and as such is under heavy development. The API's have begun to stabilize, though the process is not yet sufficiently documented, and there may be a few rough edges (particularly with generation of initial sampler states, autocorrelation issues, and detection of invalid model specifications). So far, the software has been most thoroughly tested on Linux platforms (Ubuntu/Mint), though it has also been ported to Windows. In principle, it should be usable on OSX, though I have been unable to test this and the installation is likely to be a pain. I'm happy to respond to questions and provide suggestions here.

libspatialSEIR

libspatialSEIR is a C++/OpenCL based framework for fitting Bayesian spatial SEIR and SEIRS compartmental epidemic models. The model class implemented by the software is described in the recently submitted manuscript: "An Empirically Adjusted Approach to Reproductive Number Estimation for Stochastic Compartmental Models: A Case Study of Two Ebola Outbreaks" (Brown, Oleson, Porter 2015), for which a preprint manuscript and analysis examples are available in the associated manuscript companion repository
The numerical heavy lifting is all done in C++, with the optional use of OpenCL calls to multi-core CPU's and GPU's to accelerate computation. This latter option is likely to be useful only to experts using very large data sets, and so is disabled by default. The primary interface is provided via the included R package, which uses the Rcpp library to access the lower level C++ API.

The R and C++ API's for libspatialSEIR are under heavy development, so documentation is a bit limited. For an introduction to the underlying C++ structure, doxygen class summaries are available. Additional tutorials and documentation are under development.

Additional example R code is available in the /scripts directory on the master branch.

Installation

The most up to date installation information is available on the libspatialSEIR wiki So only Linux and Windows are supported, but OSX support is planned (and might work already).