PaCAL - ProbAbilistic CALculator - Installation instructions

Requirements:

The following packages are required:

Python programming language version 2.7 (not 3.x)
NumPy numerical library
Matplotlib plotting package
Sympy symbolic computation package (for PaCAL version 1.1 or later)
SciPy Python scientific library

The following packages are recommended:

Cython if you want to compile the code yourself

Downloading and installation:

PaCal is available at PyPI so you can just do the usual

  easy_install pacal

If you are a Windows user, simply use the provided Windows installer. To install the package manually download PaCal at SourceForge, unpack the tar.gz file and type

  python setup.py install

just as with any other Python package.

Using PaCAL

To start using the package, run Python and type

from pacal import *

N = NormalDistr()
N.plot()
show()

N.cdf(0)          # should produce 0.5
N.summary()       # prints summary: mean, variance, etc.

You can compute with probability distributions just as with standard variables (independence is always assumed):

C = N / N         # Cauchy distribution
C.summary()

L = N*N + N*N     # Laplace distribution
L.plot()
show()

Many more examples are available with the package in the examples/ directory.

Enabling parallel computation

Since version 1.5 PaCAL supports parallel computation through the multiprocessing package. To enable set

params.general.parallel=True

at the beginning of your file or in the package sources. You may encounter problems with multithreaded linear algebra libraries. It is best to set the number of threads they use to one, e.g. for OpenBLAS use

export OPENBLAS_NUM_THREADS=1