At the moment the fit! checks whether flux is an argument or else, it tries to apply a more specific fit.
I think this should be improved. Basically what we want is
capture the standard cases (ols, ridge, ...) and have specific functions potentially interfacing with packages solving these cases specifically
if it's not a standard case, some basic checks should be applied (e.g.: is it differentiable, what form it has etc) and try to apply an appropriate multi-purpose optimisation solver.
this also applies for FLUX btw which can't deal with arbitrary loss functions.
At the moment the
fit!checks whetherfluxis an argument or else, it tries to apply a more specific fit.I think this should be improved. Basically what we want is
this also applies for FLUX btw which can't deal with arbitrary loss functions.