Problem Interface

Problem Interface

This page defines the common problem interface. There are certain rules that can be applied to any function definition, and this page defines those behaviors.

In-place vs Out-of-Place Function Definition Forms

Every problem definition has an in-place and out-of-place form, commonly referred throughout DiffEq as IIP (isinplace) and OOP (out of place). The in-place form is a mutating form. For example, on ODEs, we have that f!(du,u,p,t) is the in-place form which, as its output, mutates du. Whatever is returned is simply ignored. Similarly, for OOP we have the form du=f(u,p,t) which uses the return.

Each of the problem types have that the first argument is the option mutating argument. The DiffEqBase system will automatically determine the functional form and place a specifier isinplace on the function to carry as type information whether the function defined for this DEProblem is in-place. However, every constructor allows for manually specifying the in-placeness of the function. For example, this can be done at the problem level like:

ODEProblem{true}(f,u0,tspan,p)

which declares that isinplace=true. Similarly this can be done at the DEFunction level. For example:

ODEFunction{true}(f,jac=myjac)

Type Specifications

Throughout DifferentialEquations.jl, the types that are given in a problem are the types used for the solution. If an initial value u0 is needed for a problem, then the state variable u will match the type of that u0. Similarly, if time exists in a problem the type for t will be derived from the types of the tspan. Parameters p can be any type and the type will be matching how it's defined in the problem.

For internal matrices, such as Jacobians and Brownian caches, these also match the type specified by the user. jac_prototype and rand_prototype can thus be any Julia matrix type which is compatible with the operations that will be performed.

Functional and Condensed Problem Inputs

Note that the initial condition can be written as a function of parameters and initial time:

u0(p,t0)

and be resolved before going to the solver. Additionally, the initial condition can be a distribution from Distributions.jl, in which case a sample initial condition will be taken each time init or solve is called.

In addition, tspan supports the following forms. The single value form t is equivalent to (zero(t),t). The functional form is allowed:

tspan(p)

which outputs a tuple.

Examples

prob = ODEProblem((u,p,t)->u,(p,t0)->p[1],(p)->(0.0,p[2]),(2.0,1.0))
using Distributions
prob = ODEProblem((u,p,t)->u,(p,t)->Normal(p,1),(0.0,1.0),1.0)

Lower Level __init and __solve

At the high level, known problematic problems will emit warnings before entering the solver to better clarify the error to the user. The following cases are checked if the solver is adaptive:

If there is an exception to these rules, please file an issue. If one wants to go around the high level solve interface and its warnings, one can call __init or __solve instead.

Modification of problem types

Problem-related types in DifferentialEquations.jl are immutable. This helps, e.g., parallel solvers to efficiently handle problem types.

However, you may want to modify the problem after it is created. For example, to simulate it for longer timespan. It can be done by the remake function:

prob1 = ODEProblem((u,p,t) -> u/2, 1.0, (0.0,1.0))
prob2 = remake(prob1; tspan=(0.0,2.0))

A general syntax of remake is

modified_problem = remake(original_problem;
  field_1 = value_1,
  field_2 = value_2,
  ...
)

where field_N and value_N are renamed to appropriate field names and new desired values.