Reject nonsymmetric Wasserstein covariance matrices

[FlorianPfaff]

Summary

• Reject materially nonsymmetric covariance and extent matrices before Gaussian/extent Wasserstein evaluation.
• Keep the existing numerical-noise behavior by allowing relative asymmetry up to 1e-12, then symmetrizing only that tiny residual.
• Add focused regressions for both covariance arguments and both extent arguments.

Bug

_as_covariance_matrix() currently calls _symmetrize() before validating positive semidefiniteness. As a result, an invalid input such as

np.array([[2.0, 1.0], [0.0, 2.0]])

is silently replaced by [[2.0, 0.5], [0.5, 2.0]], which is positive semidefinite, and a finite Wasserstein distance is returned for a matrix that was never a valid covariance matrix.

This is also inconsistent with the NEES/NIS covariance-stack validation, which already rejects material asymmetry.

Fix

Validate covariance symmetry before _symmetrize() using the same relative 1e-12 asymmetry policy already used for covariance stacks. Tiny floating-point skew remains tolerated and is still removed by _symmetrize(); material asymmetry now raises a ValueError naming the offending argument.

Validation

• Regression coverage added to tests/test_metrics_wasserstein_validation.py.
• Branch is based on fc15bed910b324cd0619a2f2b71abc887bc09747 and is 2 commits ahead / 0 behind that base.
• Diff is limited to 13 production additions and 19 test additions across 2 files.
• Full repository CI is left to GitHub Actions because this execution environment cannot clone/install the repository dependencies.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Max errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no		no	7.25s
✅ JSON	prettier	7	0	0		0	0.65s
✅ JSON	v8r	7		0		0	5.13s
✅ MARKDOWN	markdownlint	68	0	0		0	2.16s
✅ MARKDOWN	markdown-table-formatter	68	0	0		0	0.76s
✅ PYTHON	black	1983	10	0		0	68.0s
✅ PYTHON	isort	1983	20	0		0	2.93s
✅ REPOSITORY	betterleaks	yes		no		no	1.25s
✅ REPOSITORY	checkov	yes		no		no	37.07s
✅ REPOSITORY	git_diff	yes		no		no	0.08s
✅ REPOSITORY	secretlint	yes		no		no	106.99s
✅ REPOSITORY	syft	yes		no		no	2.82s
✅ REPOSITORY	trivy-sbom	yes		no		no	2.53s
✅ YAML	prettier	11	0	0		0	0.61s
✅ YAML	v8r	11		0		0	18.61s
✅ YAML	yamllint	11		0		0	1.12s

Notices

⚠️ Your configuration references items that have been removed from MegaLinter and are ignored: MAKEFILE_CHECKMAKE. See Removed linters to find their replacements.

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@10.0.0 --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_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository