FAQ

How does pardax compare with diffrax?

diffrax is a larger, more general JAX-based differential equation library, while pardax is intentionally minimal and modelled on scipy.integrate.solve_ivp.

Some of the main differences between these two libraries are:

• Scope: diffrax supports ODEs, SDEs (stochastic), and CDEs (controlled), while pardax focuses on ODEs and PDEs only
• Solvers: diffrax currently offers a larger catalogue of solvers than pardax, however both are composable and designed to support user-defined solvers
• Stepping: diffrax has an adaptive step size controller, while pardax uses a fixed step size or a user-supplied state-dependent callback (e.g. a CFL condition)
• Spectral methods: pardax has built-in support for spectral methods, which diffrax doesn't have at the time of writing