RNG in DiscreteProblem solver

[sdwfrost]

Hi everyone, I have a DiscreteProblem that uses a random number generator, and I’d like to make it reproducible. In the solver for SDEProblems, I can specify a seed (if not an RNG) - do I need to compose a new init/step! method to maintain a specific RNG?

[ChrisRackauckas]

Is this supposed to be a state space problem?

[sdwfrost]

It’s basically this model (see code at end)

Tips on (sort of) creating a new Problem type in SciML, derived from an existing type

Is this code cleaner? Is composition the best way here, rather than use traits e.g. via SimpleTraits.jl? I mostly need the types for the time and the state, so the typing should be sufficient. What are the scenarios for using the isinplace type? # TODO: # Add MarkovSystem # Generalize to integer steps # Write integrator struct # Write init and step methods import SciMLBase: DiscreteProblem, ODESolution import OrdinaryDiffEq: solve, FunctionMap import Random: AbstractRNG import Plots: plot str…

[ChrisRackauckas]

Isn’t that the same as what’s handled in DifferenceEquations?

https://sciml.github.io/DifferenceEquations.jl/dev/