Refactor classification to have additional keyword arguments

[oameye]

No description provided.

[jkosata]

Why is this needed? The current paradigm is that whenever solutions are manipulated, a function is constructed which takes the variables in a fixed order. This function is constructed internally (hence no kwargs). You can also construct it yourself, but in this case why kwargs if these cannot change?

[oameye]

I need it for the classification of the parametrons. I classify the solutions through the normal transformation matrix which i need to give to the classification function.
With kwarg I do not have to change the classification function every time I change the topology of the network.
I think it is a nice feature. But if you think it can affect performance?

[jkosata]

I checked and kwargs should have the same performance (not sure though).
Why not just define your function with the matrix you need and pass it to classify_solutions! ? Having a construct where a constant kwarg is fed into a function you yourself just defined seems redundant.

[jkosata]

Ad performance: I checked a simple case but there have been issues with splatting (...) before (https://discourse.julialang.org/t/why-is-the-splat-operator-slow/77928) , I think this construct just opens more ways to mess up

[oameye]

Closed in favor of using anonymous functions, e.g.,
classify_solutions!(result_ωλ, soln -> is_symmetric(soln; sym_basis=sym_basis), "symmetric")