benchopt/benchmark_logreg_l1

If lmbd is parametrized as a function of lambda max, some figures in the benchopt doc are wrong

Closed this issue · 1 comments

mathurinm commented

Figure generated by benchopt example. It's lambda = lambda_max so it should converge immediately. Did we forget a scaling by 2 in logreg lambda_max ?

image

mathurinm commented

The figure in benchopt doc is for a L2 penalization, not L1