FFTL — Generalised FFTLog for Python#
The FFTL package for Python contains a routine to calculate integral transforms of the type
for arbitrary kernels \(T\). It uses a generalisation of the FFTLog [2] method of Hamilton [1] to efficiently compute the transform on logarithmic input and output grids.
The package supports any array implementation such as Numpy, JAX, Torch, dask,
etc. as long as it provides a reasonable set of standard Array API functions
(e.g. log(), exp(), linspace(), etc.) and Fast Fourier Transforms
(fft.rfft() and fft.irfft()).
Besides the generalised FFTLog algorithm, the package also provides a number of standard integral transforms implemented using SciPy.
Installation#
Install with pip:
pip install fftl
For development, it is recommended to clone the GitHub repository, and perform an editable pip installation.
The core package only requires numpy.
The standard integral transforms in the fftl.scipy module additionally
require scipy. You can install FFTL with these additional dependencies
using pip install fftl[scipy].
Usage#
The core functionality of the package is provided by the fftl module.
The fftl() routine computes the generalised FFTLog integral
transform for a given kernel.
For convenience, a number of standard integral transforms are implemented in
the fftl.scipy module.