Quentin Bertrand

I was wondering if we could do a generic l1 reweighting for an penalty with weights arguments, this way could think of reweighted Huber or reweighted lad lasso

While we are discussing the API, it seems that instantiating `HeldOutMSE` with `idx_train` and `idx_val` is counter intuitive. What do you think `HeldOutMSE` should take as argument? @agramfort @josephsalmon @Klopfe...

> It should be HeldOutMSE(None, None)? Yes you are right! thx!

Btw I think I should replace 'WeightedLasso val' by 'WeightedLassoCV' wdyt?

Hello @hermanhmchan , Thanks a lot for your interest! Could you tell us more about how important/useful this option is in your community? (maybe @Klopfe or @PABannier knows). Would you...

We currently implement a [Cox model](https://contrib.scikit-learn.org/skglm/auto_examples/plot_survival_analysis.html#sphx-glr-auto-examples-plot-survival-analysis-py). I was wondering how important is the `strata` option: it seems to be some kind of [group weight for the Cox datafit](https://stats.stackexchange.com/questions/256148/stratification-in-cox-model). If this...

I just added a way to load the design matrix in a sparse scipy format from `magenpy`. Would show us how to load `X.T y` from `magenpy` @shz9 ?

Thanks a lot for the code snippet! > Would it be possible to have the interface take` X.T X` instead of `X` in some instances? This should take more time...

This might be due to something else, but as far as I remember, our current implementation of Prox-newton might be restricted to convex penalties: there are [tricks in the line-search](https://github.com/scikit-learn-contrib/skglm/blob/ccc634487c28dd9db48a2bd7ef2f21f8092d7251/skglm/solvers/prox_newton.py#L352)...

Let us know if you need help @fsaforo1 !