skglm
skglm copied to clipboard
FEAT add compute_alpha_max function
Popped up in #251 and several other times. We have penalty.alpha_max for some penalties; having a function which does not require the gradient, but takes datafit, penalty, X and y would be helpful.