I have a problem which could be written (not exactly) as
\dot u = \min_w f(u,w)
Naturally, I would like use some previous value of \dot u as the starting point for the optimization problem on the r.h.s. — I might even like to store some additional data from the previous minimization (such as the solution of a dual problem) to speed up the optimization.
What’s the best way to do this? And can I rely on du
being unchanged during calls to my r.h.s. function (in solve
from DifferentialEquations.jl)?
EDIT: The minimization is subject to some constraints, so this is not entirely trivial.