[AIR] Add KBinsDiscretizer

[Yard1]

Signed-off-by: Antoni Baum [email protected]

Why are these changes needed?

K-Bins discretization is a common preprocessing step. This PR adds two new preprocessors - CustomKBinsDiscretizer and UniformKBinsDiscretizer. The former uses user-defined bin edges and in the latter, the user provides a desired number of bins and uniform-width bins are created automatically. Both of those preprocessors are essentially thin wrappers around pandas.cut.

In the future, a quantile-based K-Bins discretizer should be added, as that is also commonly used. It is more complex to implement, though. For the time being, users can calculate the quantiles themselves and pass them to CustomKBinsDiscretizer.

Related issue number

Closes #28301

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[bveeramani]

Would a more-general Discretizers be better? I see sklearn does something similar in their user guide https://scikit-learn.org/stable/modules/preprocessing.html#discretization

[Yard1]

Sure, makes sense!

[bveeramani]

Overall LGTM. Left some minor suggestions and questions.

Could you see if you can reproduce my inconsistent output? https://github.com/ray-project/ray/pull/28389/files#r967751046