Meekail Zain

PERF Implement `PairwiseDistancesReduction` backend for `KNeighbors.predict`

4

#### Reference Issues/PRs Fixes #13783 Resolves #14543 (stalled) Relates to #23721 Relates to #22587 #### What does this implement/fix? Explain your changes. Implements a `PairwiseDistancesReduction` backend algorithm for `KNeighbors.predict`. ####...

MAINT Update the name scheme of the `PairwiseDistancesReduction` backend

1

#### Reference Issues/PRs Fixes #24050 #### What does this implement/fix? Explain your changes. Updates names as per proposal two in #24050 #### Any other comments? - For consistency, changed `*RadiusNeighborhood*`...

ENH: Add support for structured arrays and dtypes in assertion comparison functions

4

## Related Issues/PRs Fixes #21539 ## Changes Adds support for recursively traversing structured arrays for in-depth comparison. ## Comments

Include `HDBSCAN` as a sub-module for `sklearn.cluster`

8

## Body #### Reference Issues/PRs Closes #14331 #### What does this implement/fix? Explain your changes. Integrates the excellent work done at https://github.com/scikit-learn-contrib/hdbscan and includes it as a sub-module of `sklearn.cluster`...

ENH allow for specifying CPU features to enable via `NPY_ENABLE_CPU_FEATURES` environment variable

11

## Related issues/PRs Fixes #22058 ## What does this PR add? This PR enables the usage of a `NPY_ENABLE_CPU_FEATURES` environment variable whose function is described in #22058. Note that the...

BUG: Build fails when trying to set `--cpu-baseline` to anything above `POPCNT`

1

### Describe the issue: When trying to build non-standard baseline, the build fails stating that it cannot find certain symbols. Specifically it fails for `SSE42, FMA3__AVX2`. ### Reproduce the code...

FIX: Updated dtype resolution in `_stack_along_minor_axis`

4

#### Reference issue Fixes #16569 #### What does this implement/fix? Updated dtype resolution in `_stack_along_minor_axis` to check if the output of the stack operation would require `np.int64` indices. Adds test....

RidgeCV doesn't allow `alpha=0`

7

### Describe the bug RidgeCV doesn't allow any alphas to be 0, despite the underlying `Ridge` linear model allowing such behavior. ### Steps/Code to Reproduce ```python from sklearn.datasets import load_diabetes...

Updated FAQ to include section on CUDA library initialization failures

2

CUDA library load/initialization errors have been fairly commonly reported in the past, and although error messages have improved on the XLA side, it is worth documenting an explanation and recommended...

Improves warning message for occasional CUDA memory error

1

Provides a possible solution to an XLA error which can occur when there is too little device memory to successfully create a `StreamExecutor`. See also https://github.com/google/jax/pull/19604