Malte Londschien comments

Results 30 comments of


                                            Malte Londschien

Use dtype dependent precision

Can we merge this and #865 and then work on the convergence issues. For me, hotfixing https://github.com/Quantco/tabmat/pull/408 locally already gave me a huge improvement in convergence.

How to set `max_alpha` in `GeneralizedLinearRegressorCV`?

Thanks! For `0 < l1_ratio < 1`, is this "the level of alpha required to get all coefficients to be zero had `l1_ratio=1`" or "the level of alpha required to...

How to set `max_alpha` in `GeneralizedLinearRegressorCV`?

For future reference: > For 0 < l1_ratio < 1, is this "the level of alpha required to get all coefficients to be zero had l1_ratio=1" or "the level of...

Can't make undo work

Thanks for the great work @slotThe! I just installed the `RemarkableLamyEraser` and see the same as @ericwass. Erasing works, undo does not.

Feature names aren't set if fit on `polars.DataFrame`

Thanks @stanmart! I'll close the issue.

Feature names aren't set if fit on `polars.DataFrame`

Can you reference the upstream issue in formulaic?

Update deps

Blocked by biosphere, which is blocked by https://github.com/paulkernfeld/ndarray-csv/issues/17.

I cannot replicate LightGBM's `refit` behaviours when using a custom loss function

If I set `boost_from_average: False` and don't use an `init_score`, I can replicate LightGBM's behaviour with the custom loss function. Also, refitting on the same data does not change the...

I cannot replicate LightGBM's `refit` behaviours when using a custom loss function

> I noticed you work at the same place that @neNasko1 did / does... I no longer work at QuantCo. I am currently pursuing a PhD. > are you working...

Update rand requirement from 0.8 to 0.9

xref https://github.com/rust-ndarray/ndarray/issues/1476