Please could you let me know if it is possible to compute the sensitivity of functionals within dolfinx in a similar manner as within dolfin-adjoint?

Many thanks

Please could you let me know if it is possible to compute the sensitivity of functionals within dolfinx in a similar manner as within dolfin-adjoint?

Many thanks

I don’t think dolfin adjoint supports dolfinx (yet). However depending on what you are trying to do, simply using ufl to define expressions of the sensitivities might be enough.

Thanks for the reply Calva.

Is there an equivalent of compute_gradient(J, Control()) in ufl? Or do you mean deriving and solving the adjoint equation manually?

You would have to derive it yourself. What I think @Calva means is that for quite alot of problems, consisting of only one PDE, deriving the adjoint equation is quite straightforward using ufl.derivative

Thanks Dokken.

Do you have an estimate for the release of dolfin-adjoint for dolfinx?

I do not. dolfin-adjoint is an independent project, initially created by Patrick E. Farrell, David A. Ham, Simon W. Funke and Marie E. Rognes and its newer version mainly created by @sebastkm.

It depends on what research grants one can get for redeveloping the software for DOLFINx, and who is willing to participate in this.

Thanks Dokken - this makes sense. I hope it is redeveloped for DOLFINx, it is a fantastic package.

Does anyone know if there’s been any progress towards dolfin-adjoint supporting DOLFINx?

I’m currently writing up a proposal for this. Let’s hope it gets funded.

2 Likes

Oh that’s fantastic news! Good man!! And best of luck.