Simpact Purple is a module for agent-based simulations of HIV and other sexually transmitted diseases and is part of a suite of HIV modeling tools known as Simpact. It is a useful tool for public health officials and epidemiologists which implements a parallelized algorithm based on queue theory.
Clone this repository:
git clone https://github.com/seanluciotolentino/SimpactPurple.git
And install it:
python setup.py install
Run the test script to make sure that everything worked out okay:
python bin/TestRuns.py
Examples of how to run simulations are in the bin folder. Each file contains as short description of what the script is doing. Additional documentation on how the algorithm works and how to set variables yourself can be found [here] (http://pythonhosted.org/SimpactPurple/).
Parallelism is implemented with the multiprocessing module for running on a single node (i.e. running on your laptop).
If you want to run distributed versions, scripting is more involved. First you'll need to have [open-mpi] (http://www.open-mpi.org/) and [mpi4py] (https://pypi.python.org/pypi/mpi4py) installed.
The algorithm is the same, the difference now is that openMPI handles the creation of extra processes (as opposed to the multiprocessing module). To run a single community simulation on a cluster environment you would run
mpirun -n 16 python my_script.py
This allocates 16 slots for your process. The script my_scripy.py should test for
the rank of the process and assign the nodes role accordingly (see documentation
for mpi4py). An example of how to set up a distributed
script is in bin\ApproximateBayesianComputation\ABCParameterInference.py.
Additionally bin\MigrationExploration.py is an example of running a simulation
with migration.
The most complete description of the algorithm is S. Lucio Tolentino's dissertation, hosted [here] (http://ir.uiowa.edu/cgi/viewcontent.cgi?article=5516&context=etd).
Chapter 3 details the single node parallel model. Chapter 4 details the distributed version. And Chapter 5 details the work on migration.