BayesDecider

Simulation of decision-making bayesian agents; analysis (mainly graphs) of results.

The simulations themselves are done in a C++ program, which is wrapped inside a Python extension module to ease running.

Organisation

• src/ — C++ library (build with SCons)
• decider/ — Python utilities
• pydecider.cpp — Python extension module, links with C++ library (build via setup.py)

Other files are currently obsolete but I haven't still taken the time to decide what to do with them ;)

Build recipe

Requirements

• g++
• SCons for building the C++ library
• Python 3 (I use 3.4 but I'm not sure what the bare minimum is; I think it is 3.2)
• numpy for some data manipulation and analysis

Instructions

• Run scons to build the C++ library.

• Run python setup.py install --user to install the Python utilities (runners and some data manipulation and analysis tools) at your local Python repository (no need for sudo)

  • You can also use pip install --user . to install via pip — you can later uninstall using pip uninstall BayesDecider

Et voilà! You can use it in any Python console; look for the decider.runners module.

from decider.runners import *
simulate(...)