InitializationRoutine

class probnum.diffeq.odefilter.init_routines.InitializationRoutine(*, is_exact, requires_jax)

Bases: ABC

Initialization routines for a filtering-based ODE solver.

One crucial factor for stable implementation of probabilistic ODE solvers is starting with a good approximation of the derivatives of the initial condition 1. (This is common in all Nordsieck-like ODE solvers.) For this reason, efficient methods of initialization need to be devised. All initialization routines in ProbNum implement the interface InitializationRoutine.

References

1

Krämer, N. and Hennig, P., Stable implementation of probabilistic ODE solvers, arXiv:2012.10106, 2020.

Attributes Summary

is_exact	Exactness of the computed initial values.
requires_jax	Whether the implementation of the routine relies on JAX.

Methods Summary

__call__(*, ivp, prior_process)

Call self as a function.

Attributes Documentation

Parameters

• is_exact (bool) –

• requires_jax (bool) –

is_exact

Exactness of the computed initial values.

Some initialization routines yield the exact initial derivatives, some others only yield approximations.

requires_jax

Whether the implementation of the routine relies on JAX.

Methods Documentation

abstract __call__(*, ivp, prior_process)

Call self as a function.

Parameters

• ivp (InitialValueProblem) –

• prior_process (MarkovProcess) –

Return type

RandomVariable