we really need to check code for metrics in terms of "sample_weights" and normalization.

[berndbischl]

should be carefully read, but also unit tested in terms of behavior.
maybe test sse1(w1) + sse2(w2) = sse2(w)

[berndbischl]

also not sure why we use "sample_weights" and not simply weights....

[tdhock]

weights (without sample) is frequently used in the neural network literature, to refer to the learned parameters