Michael Mayer comments

Results 161 comments of


                                            Michael Mayer

ENH Add Friedman's H-squared

> Yep. I might need some time to catch up with the topic and the code. But this is on my shortlisted PR for the next release. Fantastic! Fresh eyes...

Option to specify random effects?

ML workflows with clustered data are a delicate thing. Using a clean train/test split (grouped split on subject) and then evaluating the model on the test data is often a...

python3: regression for unique on dtype=object arrays with varying items types (Trac #2188)

Same words as @mcgfeller . I would even take a slowish non-sorted version.

Interactions

I 👍. The question is: is it efficient? (Interactions with dummies generate many 0). And: is it safe to load a serialized model and use it to predict on unseen...

Interactions

Wow, thanks a lot for the explanations. Really looking forward to this!

LightGBM categoricals and shap_interactions

@wkondrusiewicz In my example above, there are no pandas structures at all. So I currently do not understand the suggested workaround. Hints are very welcome :-). The problem is with...

LightGBM categoricals and shap_interactions

I think I have tracked the problem: SHAP turns the original tree ensemble into its own tree structure by calling `TreeEnsemble`. There, we find the following tree transformation block (multiple...

LightGBM categoricals and shap_interactions

@p-schaefer : see also https://github.com/microsoft/LightGBM/issues/6058

Add required variables to the `Formula` class

This would indeed be very handy, thx.

[R-package] lightgbm::lgb.model.dt.tree() error caused by lgb.dump() error with large models

Note on the status: We have finished https://github.com/microsoft/LightGBM/issues/6397, and I have started to draft a solution for this issue, but it is not yet in the state I want it...