[Vp-reproduce-subgroup] [Vp-integration-subgroup] [EXT] Credibility “for EACH USE”Re: White paper revision

[William Waites]

I wrote:

> φM(x) = φA(φB(x)) + O(h) 

That should be O(h²) i.e. it’s a first order integrator, much like Euler’s method. With some work, you can build integrators to arbitrarily high order through splitting and composition like this. That you can prove the order of these integrators means that we can say something useful about composition in this context. Normally, in numerical analysis, this kind of exercise is talked about in terms of building “custom integrators” in an application specific way, somewhat analogous to what we’ve been talking about doing with composing modular models.