Apply new validation to cuml.random_projection

[jcrist]

Fixes #8003.

Note that this fixes a small bug in fit, bringing the meaning of convert_dtype in line with the rest of cuml.

[jcrist]

This test and behavior was incorrect and doesn't match our intended behavior of convert_dtype. I view the change here as a bugfix. Also note that I wrote the code (and test) a few months ago, so I know this wasn't intentional for a specific user need, just a buggy implementation of a standard kwarg.

[coderabbitai]

📝 Walkthrough

Walkthrough

This pull request refactors the random projection module to use centralized input validation (check_inputs) for both fit and transform methods. It updates dtype handling to preserve input dtypes (float32/float64) throughout the pipeline, modifies sparse matrix construction to use cupyx.scipy.sparse, and adjusts test expectations and sklearn compatibility configurations accordingly.

Changes

Cohort / File(s)	Summary
Input Validation Refactoring python/cuml/cuml/random_projection/random_projection.py	Replaces manual fit/transform input handling with check_inputs validation. Changes reflection reset semantics from reset=True to reset="type", removes manual dtype logic tied to convert_dtype, validates/coerces X to float32/float64, and updates sparse matrix construction to use cupyx.scipy.sparse. Simplifies sparse handling in output generation.
Test Suite Updates python/cuml/tests/test_random_projection.py	Renames test_components_dtype to test_components_and_output_dtype and updates validation to verify that projection matrix and fit_transform preserve input dtype. Removes prior expectations for default float32 outputs and dtype alteration via convert_dtype=False.
Sklearn Compatibility Configuration python/cuml/tests/test_sklearn_compatibility.py	Removes xfail decorators for complex-data, object-dtype, empty-data, and NaN/inf checks on GaussianRandomProjection and SparseRandomProjection, while retaining other compatibility xfail configurations.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 40.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'Apply new validation to cuml.random_projection' directly and clearly describes the main change: applying new validation to the random_projection module.
Linked Issues check	✅ Passed	The PR fully addresses issue #8003's objective to update cuml.random_projection to new input validation by implementing centralized check_inputs validation/coercion in fit/transform.
Out of Scope Changes check	✅ Passed	All changes are directly related to implementing the new input validation in random_projection module and updating tests accordingly; no out-of-scope modifications detected.
Description check	✅ Passed	The pull request description directly references the related issue (#8003) and briefly explains the intent: applying new validation to random_projection and fixing a convert_dtype bug.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@python/cuml/cuml/random_projection/random_projection.py`:
- Around line 89-97: The fit() path currently calls check_inputs(...) with
accept_sparse=True which allows any sparse format but transform() only accepts
CSR/CSC; update the call in random_projection.py (inside the fit method where
check_inputs is invoked) to restrict accept_sparse to the same formats as
transform(), e.g. accept_sparse=("csr", "csc"), so fit(), transform(), and
fit_transform() accept the same sparse matrix types and avoid inconsistency.

In `@python/cuml/tests/test_random_projection.py`:
- Around line 170-177: The test test_components_and_output_dtype only exercises
dense inputs; update it to parametrize over sparse inputs as well (e.g., add a
parameter like "sparse" or "input_format") and when sparse=True convert X to a
scipy.sparse CSR/CSC matrix before calling model = cls(...); ensure you still
call transformed = model.fit_transform(X) and assert both
model.components_.dtype == dtype and transformed.dtype == dtype for the sparse
branch so the sparse conversion/validation path is covered; reference the
existing symbols random_array, classes, test_components_and_output_dtype,
model.fit_transform, model.components_, and transformed when locating where to
add the new param and assertions.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 72da559c-8696-4258-921f-557bd8c006fa

📥 Commits

Reviewing files that changed from the base of the PR and between 42f57b5 and f41be04.

📒 Files selected for processing (3)

• python/cuml/cuml/random_projection/random_projection.py
• python/cuml/tests/test_random_projection.py
• python/cuml/tests/test_sklearn_compatibility.py

💤 Files with no reviewable changes (1)

• python/cuml/tests/test_sklearn_compatibility.py

[viclafargue]

Thanks! Just a bunch of questions to draw attention to the input processing in the fit function. Pre-approving should be good to go once solved.

[jcrist]

/merge