Closed JakobAsslaender closed 1 year ago
Ah ok. It seems that MRIReco.jl calls the proximal operator with all kinds of "junk" keyword arguments that have no impact on this proximal operator. As long as we keep it this way, there might be no way around using kwargs
as a garbage collector. We could change this with a complete overhaul of the RegularizedLeastSquares.jl interface that @tknopp, I, and others discussed a while back. Essentially, by removing the generic wrapper constructors and replacing them with regularization-specific constructors that inherit from an abstract regularizer class.
Long story short: I guess we should keep the current master version as is and revisit this idea when someone has the time to overhaul this package. Does that make sense?
I sketched an interface w/o the
kwargs
arguments. @tknopp, @JAgho: Does that work?