Fix Random Projector magnitudes

[andrewdalpino]

Transformers/GaussianRandomProjector.php:120, SparseRandomProjector.php:82 — [upstream-check] missing the JL 1/sqrt(k) normalization (preserves ratios, not absolute magnitude).

• GaussianRandomProjector — matrix scaled by ÷ √k at fit (src/Transformers/GaussianRandomProjector.php:122)
• SparseRandomProjector — dHat changed to 1/√(density · k) (src/Transformers/SparseRandomProjector.php:82)

[Copilot]

🟡 Changes recommended

SparseRandomProjector::fit() can still deterministically divide by zero when sparsity is 1.0 (density = 0), producing invalid projection matrices instead of failing fast.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR adjusts the random projection matrix scaling in both GaussianRandomProjector and SparseRandomProjector to include the Johnson–Lindenstrauss (1/\sqrt{k}) normalization so that projected vectors better preserve absolute magnitude (not just relative ratios).

Changes:

• Scale GaussianRandomProjector’s random matrix by 1 / sqrt(dimensions) during fit().
• Update SparseRandomProjector’s distribution magnitude (dHat) to 1 / sqrt(density * dimensions).
• Revise both projector tests to validate approximate magnitude preservation instead of asserting exact projected coordinates.

File summaries

File	Description
src/Transformers/GaussianRandomProjector.php	Applies JL normalization by dividing the Gaussian matrix by sqrt(k) at fit time.
src/Transformers/SparseRandomProjector.php	Updates sparse projection scaling to include dimensions in the normalization factor.
tests/Transformers/GaussianRandomProjectorTest.php	Replaces exact-vector assertion with a magnitude-preservation check over a dataset.
tests/Transformers/SparseRandomProjectorTest.php	Replaces exact-vector assertion with a magnitude-preservation check over a dataset.

Review details

• Files reviewed: 4/4 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.