Julien Jerphanion

> Can you elaborate on this later aspect? Yes, can the axioms (listed in section 2.6 of [the original paper](https://arxiv.org/pdf/1908.08474.pdf)) be realized in practice? > If there are issues regarding...

> @samronsin @jjerphan may I ask about the next steps here? If I correctly followed the issues, this would be the basis to have random forests with monotonic constraints. #18982...

> Actually @mayer79 this PR already implements monotonicity constraints for trees ensembles (including Random Forests), so this would be it, no next step. In this case, can this PR title...

03ee5f4 seems to have fixed performance regressions: asv benchmarks ``` → asv continuous -b RandomForest -b GradientBoostingClassifier main monotonic-trees · Creating environments · Discovering benchmarks ·· Uninstalling from conda-py3.9-cython-joblib-numpy-scipy-threadpoolctl ··...

@samronsin: thanks for pursuing your work. :handshake: Do you need help with the recent refactoring changes made on the heaps?

Thanks for this work, @cmarmo! > Hello, I'm unable to the debug on Windows ... perhaps someone might have a look to the last failing Windows test? Thanks! It's weird...

A merge and _voilà_: all green ! :heavy_check_mark: :upside_down_face: More seriously, I think it's weird that this test failed before but now passes.

> @jjerphan , that would be awesome ... but ... the test does not pass during the wheel building with Windows python 3.11 and scipy 1.10 ... Ah yes, my...

> Maybe a bit radical, but do we need to support win32 on Python 3.11? Or at all? This PR does not propose adding Python 3.11 wheels for win32. scikit-learn...

Would it make sense to release wheels for other platforms than Windows first, and then resolve problems for Windows 32bit and 64bit platforms separately?