Stabilize nonadditive particle-noise weight normalization

[FlorianPfaff]

Summary

• normalize nonadditive particle-noise weights after scaling by their largest entry
• preserve valid relative probabilities when the direct floating-point sum would overflow
• retain existing validation for empty, non-finite, negative, and zero-mass inputs
• add a backend-aware regression through the public Euclidean particle-filter API

Root cause

AbstractParticleFilter.predict_nonlinear_nonadditive() validated each noise weight as finite and nonnegative, then summed the raw vector directly:

weight_sum = sum(weights)
weights = weights / weight_sum

Individually finite weights can still have an infinite sum. For example, [finfo.max, finfo.max / 2] has the well-defined ratio 2:1, but its direct sum overflows. The method consequently rejected a valid nonadditive-noise distribution before sampling.

Fix

Scale the weights by their largest positive entry before summing. The scaled entries lie in [0, 1], preserve the original ratios, and normalize to finite unit mass. An explicit empty-vector guard preserves the previous clean ValueError contract.

Validation

• confirmed the old direct sum overflows for both NumPy float32 and float64
• confirmed scale-first normalization produces [2/3, 1/3]
• added a regression that captures the probability vector passed to backend random.choice() and verifies the resulting nonadditive prediction
• Python syntax compilation passed for the modified module and regression test
• branch is two commits ahead of current main, zero behind, and changes only the particle-filter normalization path plus one focused test

The full multi-backend test matrix is delegated to GitHub Actions.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	23.33s
✅ JSON	prettier	7	0	0	0	1.08s
✅ JSON	v8r	7		0	0	5.02s
✅ MARKDOWN	markdownlint	68	0	0	0	2.05s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	1.17s
✅ PYTHON	black	1490	183	0	0	96.52s
✅ PYTHON	isort	1490	329	0	0	2.65s
✅ REPOSITORY	betterleaks	yes		no	no	2.09s
✅ REPOSITORY	checkov	yes		no	no	53.7s
✅ REPOSITORY	gitleaks	yes		no	no	15.05s
✅ REPOSITORY	git_diff	yes		no	no	0.32s
✅ REPOSITORY	secretlint	yes		no	no	57.96s
✅ REPOSITORY	syft	yes		no	no	5.99s
✅ REPOSITORY	trivy-sbom	yes		no	no	7.65s
✅ REPOSITORY	trufflehog	yes		no	no	32.81s
✅ YAML	prettier	11	0	0	0	0.75s
✅ YAML	v8r	11		0	0	12.19s
✅ YAML	yamllint	11		0	0	0.77s

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