You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The communication between multiple numerical models for a boundary problem operating within the same geometry requires to efficiently sharing information. Some of these pieces of information are relativelly trivial, such as state, time, and parameters, but others are way less straightforward, such as integration_times and, most importantly, integration_grid.
Suggested approach
Understand the problem.
Brainstorm about the problem.
Google a bit. It is hard to believe this is a novel problem.
Ask internally. Good candidates are Lourens and Johan.
Implement a solution. Here is my suggestion:
Minimal: enforce the use of the same grid ab initio for all integrators.
Intermediate: use direct interpolation, i.e.:, assume that if $I(x)$ returns the interpolated value of $x$, then $F(I(x)) \approx I(F(x))$
Excellent: use automatic differentiation for a more system-informed interpolation, i.e.:, assume that $F(I(x)) \approx F(x) + F'(x) \epsilon + O(\epsilon^2)$, where $\epsilon$ depends on the interpolation method $I$.
The text was updated successfully, but these errors were encountered:
Problem
The communication between multiple numerical models for a boundary problem operating within the same geometry requires to efficiently sharing information. Some of these pieces of information are relativelly trivial, such as
state
,time
, andparameters
, but others are way less straightforward, such asintegration_times
and, most importantly,integration_grid
.Suggested approach
The text was updated successfully, but these errors were encountered: