Stabilize MEM-RBPF particle log likelihoods

[FlorianPfaff]

Bug

MEMRBPFTracker._update_particle_weights() formed det(marginal_cov) before taking its logarithm. For valid positive-definite but very small covariance matrices, the determinant can underflow to zero in float64.

When that happened, the affected particle likelihoods were replaced with -inf. If every particle covariance underflowed, log-weight normalization fell back to uniform weights and discarded otherwise valid measurement information.

Fix

• compute the covariance log-determinant from its symmetric eigenvalues without forming the determinant;
• reject non-finite or non-positive eigenvalues before evaluating the likelihood;
• use only linear-algebra operations already exported by the NumPy and PyTorch backend facade;
• add a regression with two valid covariance scales whose ordinary determinants both underflow.

Impact

MEM-RBPF particle updates now preserve relative likelihood information for extremely concentrated valid covariances instead of silently resetting the particle weights to uniform.

Validation

• branch is 3 commits ahead of main and 0 behind;
• final diff is limited to 6 source additions/3 deletions and one 38-line regression test;
• analytical reproduction: the old determinant path yields [-inf, -inf], while the stable path produces normalized weights [2/3, 1/3];
• the implementation uses eigvalsh, isfinite, log, and reductions already supported by the MEM-RBPF backends;
• repository GitHub Actions exercise the focused NumPy regression and the broader test suite.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	23.53s
✅ JSON	prettier	7	0	0	0	1.14s
✅ JSON	v8r	7		0	0	4.91s
✅ MARKDOWN	markdownlint	68	0	0	0	1.81s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.89s
✅ PYTHON	black	1585	221	0	0	92.27s
✅ PYTHON	isort	1585	401	0	0	2.67s
✅ REPOSITORY	betterleaks	yes		no	no	2.73s
✅ REPOSITORY	checkov	yes		no	no	52.56s
✅ REPOSITORY	gitleaks	yes		no	no	15.46s
✅ REPOSITORY	git_diff	yes		no	no	0.69s
✅ REPOSITORY	secretlint	yes		no	no	63.09s
✅ REPOSITORY	syft	yes		no	no	5.65s
✅ REPOSITORY	trivy-sbom	yes		no	no	5.39s
✅ REPOSITORY	trufflehog	yes		no	no	31.77s
✅ YAML	prettier	11	0	0	0	0.75s
✅ YAML	v8r	11		0	0	12.63s
✅ YAML	yamllint	11		0	0	0.52s

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