DynODE Simulation Module

DynODE Simulation Module#

The simulation module provides utilities for solving systems of ordinary differential equations (ODEs) using the diffrax library. For more information on diffrax in DynODE check out the it’s section in the back-end libraries section

simulate Function#

simulate(
    ode: ODE_Eqns,
    duration_days: int,
    initial_state: CompartmentState,
    ode_parameters: AbstractODEParams,
    solver_parameters: SolverParams,
    sub_save_indices: Optional[Tuple[int, ...]] = None,
    save_step: int = 1,
) -> Solution

The simulate function numerically solves a user-supplied ODE system over a given number of days, returning a diffrax.Solution object containing the compartment states at each saved time point.

our simulate function is a thin wrapper around the diffrax diffeqsolve method so users dont need to learn the libraries syntax.

Parameters#

  • ode: user supplied callable ODE function with signature as follows

    • def ode(t: float, y: CompartmentState, params: AbstractODEParams) -> CompartmentGradients

  • duration_days: Number of days to simulate.

  • initial_state: Tuple of JAX arrays representing the initial state of each compartment.

  • ode_parameters: Chex parameters object for the ODE, passed through to params in the ode.

  • solver_parameters: a SolverParams object containing solver configuration, such as step size and error tolerances.

  • sub_save_indices: Optional tuple of indices specifying which compartments to save.

  • save_step: Interval (in days) at which to save the solution.

Returns#

  • diffrax.Solution: Contains the time points and compartment states for the simulation. Solution documentation

Notes#

  • All compartment states must be JAX arrays.

  • The ODE function parameter order matters must accept parameters in the order: (t, y, params).

Contents