|Prof Darren Wilkinson||Professor of Stochastic Modelling|
|School of Mathematics & Statistics|
gdagsim-03.tgz is the current release of GDAGsim. This is a C library for analysis and block-sampling of Gaussian DAG models. Full documentation and installation instructions are included with the package - see the README in the top-level directory. To use the library you will need to be reasonably familiar with programming in C, and it helps (but is not vital) if you've used the GNU Scientific Library (GSL) before. Note that you must have the GSL (>=1.0) installed and working correctly before attempting to install GDAGsim. Since there are (currently) no sparse matrix algorithms in the GSL, this version of GDAGsim also depends on the Meschach matrix library. However, no familiarity with Meschach is required.
The main change from 0.2 is a change of syntax for the stochastic simulation functions. This syntax change is annoying, but the decision to introduce it has not been taken lightly! The library is now thread-safe. This will ease the development of multi-threaded and parallel codes which use the library, and also has some other benefits.
If you use Linux, then pre-compiled binary packages are available for both the GSL and Meschach - just install the library packages and the corresponding "-dev" packages before installing GDAGsim, and everything should work fine. I hope to put together binary packages for GDAGsim for use with Debian GNU/Linux sometime in the next few months, but for now, building from source should be very straightforward.
The latest documentation is available in PDF format for on-line browsing. The theory behind GDAGsim is explained in the following paper:
gdagsim-01.tgz is an alpha release of GDAGsim, my C library for analysis and block-sampling of Gaussian DAG models.
Version 0.1 depends only on the GSL. However, this version uses dense matrix algorithms, and hence is very inefficient. It is no longer undergoing development.
gdagsim-02.tgz is the first version of GDAGsim to use sparse matrix algorithms. It is much more efficient than Version 0.1 in terms of both memory usage and CPU time. However, 0.1 and 0.2 both suffer from not being thread-safe, due to the way they handle random number streams. It is no longer being developed. Version 0.3 corrects this problem, but therefore requires a syntax change to the stochastic simulation functions.