rvmath 0.1

Quick Installation

To install rvmath, simply (soon):

$ pip install rvmath

and then simply enjoy it!

Other functionality

All rvmath objects have a few useful options and methods:

You can assign an id to the random variable:

>>> x = rvm.uniform(rvid="x")

to simplify debugging. If rvid is not given, a random string will be generated.

You can ask for dict containing all random variable names and their underlying SciPy distribution objects.

>>> x = rvm.uniform(rvid="x")
>>> y = rvm.norm(rvid="y")
>>> z = x + y
>>> dict(z.random_vars())
{'x': <scipy.stats._distn_infrastructure.rv_frozen at 0x7ff57f196220>,
 'y': <scipy.stats._distn_infrastructure.rv_frozen at 0x7ff57e5a81f0>}

You can draw values from all random variables within an object.

>>> realization = z.draw(3)
>>> print(realization)
{'x': array([0.75633395, 0.99657116, 0.26853511]),
 'y': array([-1.23407414,  0.5261816 ,  2.62764828])}

and finally you can evaluate the object for this particular realization.

>>> z.eval(realization)
array([-0.47774019,  1.52275276,  2.89618339])

This is exactly what happens when rvs is called, but it can be particularly useful for debugging, testing and evaluating subexpressions.

In certain cases you might want to predefine the size of certain random variable. You can do it using the size parameter upon construction.

>>> z = np.sum(rvm.uniform(size=(3, 3))) * rvm.norm()

Random variables without an explicit size parameters will be evaluated to the size given when calling rvs.

Together with this feature, you can build distributions that depend on others:

>>> m = rvm.uniform(size=1)
>>> w = rvm.norm(loc=m)

You can combine distribution defined size with rvs provided size

>>> m = rvm.uniform(size=(None, 3, None)).rvs(2)
>>> m.shape
(2, 3, 2)

None values will be replaced by scalars provided to rvs.

FAQ

1. Which distributions are supported by rvmath?

   All continuous distributions available in Scipy, you can read the list in the SciPy Stats docs.

2. Do you plan tu support discrete distributions?

   Yes, hopefully in the next version.

3. I have built my own distribution subclassing ``rv_continuous``, can I use it with rvmath?

   Yes, just call rvm.wrap(distro, *args, **kwargs)

rvmath is maintained by a community. See AUTHORS for a complete list.

To review an ordered list of notable changes for each version of a project, see CHANGES