Fix cancellation in Merwe sigma-point scaling

[FlorianPfaff]

What changed

• compute the Merwe covariance scale directly as alpha**2 * (n + kappa) instead of reconstructing it through n + (scale - n)
• use the direct scale consistently for sigma-point generation and weights
• reject underflowed, overflowed, non-normalizable, or non-finite weight configurations with explicit ValueErrors
• add focused regressions for cancellation, scale underflow/overflow, and float64 weight normalization limits

Root cause

The implementation first formed lambda = alpha**2 * (n + kappa) - n and then reconstructed the covariance factor as n + lambda. When the intended factor is small relative to n, the subtraction and re-addition lose significant digits through catastrophic cancellation.

For n=1, kappa=0, and alpha=2e-8, the intended scale is 4.0e-16, but the reconstructed scale is 4.440892098500626e-16. This inflates the scale by about 11.02% and the sigma-point displacement by about 5.37%.

Impact

Small but representable Merwe scaling parameters now produce the requested sigma-point spread instead of a rounded, wider spread. Parameter combinations that cannot produce finite and normalized float64 weights fail clearly rather than returning distorted weights or leaking ZeroDivisionError/OverflowError.

Validation

• reproduced the pre-fix scale and spread errors in a focused numerical harness
• verified the direct-scale calculation produces the requested ±2e-8 sigma-point offsets
• added tests/test_sigma_points_merwe_scale_stability.py
• branch is two commits ahead of main and zero behind; full repository CI is pending

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	26.78s
✅ JSON	prettier	7	0	0	0	1.57s
✅ JSON	v8r	7		0	0	3.95s
✅ MARKDOWN	markdownlint	68	0	0	0	2.18s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.71s
✅ PYTHON	black	1795	47	0	0	92.16s
✅ PYTHON	isort	1795	76	0	0	2.7s
✅ REPOSITORY	betterleaks	yes		no	no	2.49s
✅ REPOSITORY	checkov	yes		no	no	54.24s
✅ REPOSITORY	gitleaks	yes		no	no	16.66s
✅ REPOSITORY	git_diff	yes		no	no	0.37s
✅ REPOSITORY	secretlint	yes		no	no	74.56s
✅ REPOSITORY	syft	yes		no	no	5.79s
✅ REPOSITORY	trivy-sbom	yes		no	no	6.55s
✅ REPOSITORY	trufflehog	yes		no	no	31.78s
✅ YAML	prettier	11	0	0	0	0.77s
✅ YAML	v8r	11		0	0	12.04s
✅ YAML	yamllint	11		0	0	0.65s

Notices

📣 MegaLinter 9.5.0 is out! Discover the new features and security recommendations in the release announcement. (Skip this info by defining SECURITY_SUGGESTIONS: false)

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@9.6.0 --custom-flavor-setup --custom-flavor-linters PYTHON_BLACK,PYTHON_ISORT,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_GITLEAKS,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository