Evaluate forward model state variables during optimization.

[Evan Cummings]

I would like to evaluate individual elements of my objective function and state variables during optimization.

I need this to analyze the performance of my solver.

I really just need the state variable, which I could then assign to the variable used in the objective functionals that I assemble each iteration with the eval_cb function utilized by ReducedFunctional.

Also, it would be nice to have access to the adjoint variables as well, for plotting purposes and such.

Can this be done?

[Simon Funke]

What you are looking for is DolfinAdjointVariable.tape_value() - it retrieves the Function objects on dolfin-adjoint tape of its latest run.

Here is an example how this can be used:

http://www.dolfin-adjoint.org/en/latest/documentation/mpec/mpec.html

[Evan Cummings]

Thanks for this!