DNest4

DNest4 is a C++11 implementation of Diffusive Nested Sampling, a Markov Chain Monte Carlo (MCMC) algorithm for Bayesian Inference and Statistical Mechanics. You can use it in a few different ways:

• Implement your model in C++, compile it and have it run super fast.
• Implement trans-dimensional models with the RJObject template class.
• Implement your model by writing just two functions in Python, R, or Julia (this is new and undocumented - email me or take a look in the Templates directory)
• Write up your model in Python, using a BUGS-style approach see here to learn how (more documentation here).

Papers

There is a paper describing DNest4 installation and usage in the Journal of Statistical software. You might also want to read the original paper describing the Diffusive Nested Sampling algorithm itself. If you find this software useful in your research, please cite one or both of these papers. Here are the citations:

Brewer, B., & Foreman-Mackey, D. (2018). DNest4: Diffusive Nested Sampling in C++ and Python.
Journal of Statistical Software, 86(7), 1 - 33. doi:http://dx.doi.org/10.18637/jss.v086.i07

Brewer, B. J., Pártay, L. B., & Csányi, G. (2011). Diffusive nested sampling.
Statistics and Computing, 21(4), 649-656.

NOTE: The Python model building described in the paper is now removed, and there is no longer a Python package. Just a couple of helper scripts for simplicity.

Dependencies

You will need a C++ compiler that supports the C++11 standard, along with Python 3 and the Python packages NumPy and matplotlib.

Compiling the C++

You can compile the DNest4 library (libdnest4) using the Makefile in the code directory using:

cd code
make

Along with building the library this will compile all the examples, which you can run by executing main in the appropriate example directory. Then, for postprocessing, run showresults.py in each example directory. The showresults.py file makes use the Python scripts from the python directory within DNest4. I suggest adding this directory to your PYTHONPATH. You should also add the directory containing the DNest4 directory to an environment variable DNEST4_PATH. This will help some Makefiles in the example/template directories to find the correct files when compiling.

Note for Mac users:

On some Macs, g++ is an alias for clang, which is a C compiler. If this is the case for you, you'll need to edit the first line of the Makefile so that it uses clang++, which is a C++ compiler.

Alternative build process with SCons

However, you can also compile and install the library using SCons. To do this you just need to run:

scons install

By default it will attempt to install the library in /usr/local (with the library files in /usr/local/lib and the headers in /usr/local/include/dnest4), so the above command must be run as a user with root access or using sudo. To install to a different location you can instead run:

scons install --prefix <install_location>

where <install_location> is the base path for the install.

To install with GDB enabled during the library's compilation you can add the --debug-mode flag to the install command.

Currently, the Scons installation does not compile the examples or the Python library. Any additions to this installation process are welcome.

(c) 2015--2018 Brendon J. Brewer and contributors. LICENCE: MIT. See the LICENSE file for details.

This work was supported by a Marsden Fast Start grant from the Royal Society of New Zealand.