Skip to content

An example

The example showcases how to use the PyARV package.

NumPy arrays

  • We use numpy arrays, not lists.
  • The arrays have a specific datatype, in this case np.float32.
  • The function needs a preallocated location to write its results into.
Example Gaussian transformation.
import numpy as np  #(1)!
from pyarv.gaussian.approximation import Gaussian
from scipy.stats import norm

n_samples = 10_000_000  #(2)!
u = np.random.uniform(size=n_samples).astype(np.float32)  #(3)!
z_approx = Gaussian(order=1).transform(u)  #(4)!
z_exact = norm.ppf(u)  #(5)!
  1. NumPy is required for the array data type and the uniform random numbers which will be transformed.
  2. This package is designed for applications which use lots of random numbers.
  3. Our approximations require 32-bit floats.
  4. The PyARV method for transforming uniform random numbers into approximate random numbers.
  5. The regular method for transforming uniform random numbers into exact random numbers.

To see how much faster PyARV is on your system, run: 

python -m pyarv.gaussian.demos.speed
python -m pyarv.non_central_chi_squared.demos.speed