Library / Code module
Covariance intersection fusion
covariance_intersection_fusion
Fuse two mean and covariance estimates whose errors may be correlated in an unknown way, choosing the weight that minimizes the trace or determinant of the fused covariance. Use it when evidence sources may share inputs, so the independence rule would overstate certainty.
ATTRIBUTION.md3.9 KBLICENSE1.0 KBREADME.md3.2 KBatom_cli.py3.0 KBcomponent.json12.9 KBcovariance_intersection_fusion.py6.2 KBnumerics.py28.0 KBschema_lite.py8.9 KBtest_covariance_intersection_fusion.py5.4 KBtest_package.py12.0 KB
Full files in your harness.
What it takes and gives
- Runs with
- Python 3.10 or later
- Entry points
- covariance_intersection, independent_fusion, conservative_margin
- Known answers
- 10 cases its tests reproduce
Cases it covers
- duplicate does not shrink: Halving the covariance of a duplicate, as the independence rule does, counts the same evidence twice.
Limits and scope
- Limits
- Two estimates per call; fuse more by repeated calls, which is conservative but not the joint optimum. Dense matrices for small dimensions (tens). Covariance intersection is conservative by construction and can be loose when the errors are in fact independent; it does not detect inconsistent inputs or outliers, and a covariance that is itself optimistic stays optimistic after fusion.
- Guarantees
- For every omega in [0, 1] and every valid cross-covariance, the fused covariance minus the true fused error covariance is positive semidefinite. Fusing an estimate with an exact duplicate returns the same mean and covariance. The returned omega minimizes the chosen criterion over [0, 1] to about 1e-16; a flat objective returns omega = 0.5.
- Assumes
- Each input covariance is a consistent (not optimistic) description of its own estimate's error. Nothing is known about the cross-covariance of the two errors.
Details
- Kind
- Code module
- Licence
- MIT
- Declared effects
- runs commands
- Source
- github.com/alisonjieli-png/loop-engine/tools/creative_originals/evaluation_atoms/items/covariance_intersection_fusion
- Size
- 10 files, 84.5 KB
- Identity
library.supply.creative_originals.129ff1439cef30a090f27917.6f75e7c178401239- Version
6f75e7c178401239b49b754f5b1aa0aa2394ab7e74d254b1a575fab40afc5612
The version is the SHA-256 of the item your harness downloads; every file in it names its own checksum.
Where it is shown
- Statistics for comparing resultsDeterministic skills
Related components
Something wrong with this item, its rights or its safety? Report it on the support page, or from your harness after a download.