xaj 0.1.8

ODE integrator for JAX

.. warning:: This version of XAJ works only with jax and jaxlib 0.3. But jaxlib 0.3 is not available on PyPI. To install this version of XAJ, please first install jaxlib with:

.. code-block:: bash
    pip install jaxlib==0.3.25 -f https://storage.googleapis.com/jax-releases/jax_releases.html

XAJ

Ordinary differential equation (ODE) integrator compatible with Google's JAX.

XAJ implements the Runge-Kutta Dormand-Prince 4/5 pair (DP5) with adaptive step control and dense output according to the Numerical Recipes. It provides a fast and efficient way to solve non-stiff ODEs. The programming interface is designed so it feels similar to the derivative functions in JAX.

Specifically, XAJ provides a single function odeint(). Applying it to another function rhs() and the initial conditions returns the numerical solution as a callable ns, which interpolates the dense output.

from xaj import odeint
from jax import numpy as np

rhs = lambda x, y: y
x0  = 0
y0  = 1

ns  = odeint(rhs, x0, y0, 1)

x   = np.linspace(0, 5)
y   = ns(x)

The numerical integration happens in a "lazy" way, which is triggered by the extreme values of the argument of ns. Alternatively, it is possible to obtain the numerical solutions at the full steps by

xs = ns.xs
ys = ns.ys

Demos on how to use XAJ can be found in the demos directory.