randify 0.1.0

Simple probability density function estimation for existing code

Randify

Randomize your existing code with randify. Evaluate how your code reacts to randomly distributed inputs and evaluate the probability distributions of your code outputs. Randify performs Monte Carlo simulations to estimate probability density distributions of your outputs. Designed for simple syntax.

Full API documentation at https://jonasnebl.github.io/randify/randify.html

Installation

Install randify from PyPI:

$ pip install randify

Quick guide

Randify works on functions with any number and type of input and output arguments. In this quick example we have a function y1, x2 = f(x1, x2) returning the sum x1 + x2 and the product x1 * x2 of two arguments x1 and x2. For more in-depth examples check-out the Jupyter-Notebooks in examples.

def f(x1, x2):
    return x1 + x2, x1 * x2

x1 = 1
x2 = 2
y1, x2 = f(x1, x2)

Now we want to evaluate how random inputs x1 and x2 influence the results y1and y2 using randify. We can do this using two steps.

1. Define x1and/or x2as a RandomVariable. For defining the RandomVariable you need to pass a function that generates random samples of this RandomVariables. In this example, we use functions from numpy's random module, but you can also define custom functions (see the examples for that).
2. Call the randify function wrapper with the RandomVariables as arguments.

from randify import randify, RandomVariable
import numpy as np

def f(x1, x2):
    return x1 + x2, x1 * x2

x1 = RandomVariable(np.random.normal, loc=0, scale=1)
x2 = RandomVariable(np.random.uniform, low=-1, high=1)
y1, y2 = randify(f)(x1, x2)

y1and y2 are now also RandomVariables. You can calculate the resulting statistical measure like expected value or variance. To display the estimated probability distributions, you can use randify's plot_pdf function:

print(f"E[y1] = {y1.expected_value}")
print(f"Var[y1] = {y1.variance}")

from randify import plot_pdf
plot_pdf(x1=x1, x2=x2, y1=y1, y2=y2)

Documentation

Full API documentation can be found at https://jonasnebl.github.io/randify/randify.html. The documentation is generated automatically using pdoc.

$ pip install pdoc
$ pdoc randify --math

Formatting

ruff is used to format randify.

$ pip install ruff
$ ruff format