Partial correlation improved link ranks at 16-fold noise variance in two passive worm models with half the cells observed
We tested unequal independent current noise in passive electrical models of two partial adult worm maps. Each model had 180 cells. In 100 noise assignments and 90-cell observation masks per map, a 16-fold noise-variance range gave mean precision at the true link count of 0.8303 and 0.8180 for inverse covariance. Partial correlation raised these values to 0.9497 and 0.9405 and won every paired comparison. Under equal noise and half observation, it reduced mean precision. A three-cell path also gave positive partial correlation of 0.4468 between cells without a direct link under unequal noise. We provide code, source hashes, aggregate scores and checks. Exact population covariance, assumed conductance, independent white noise and known link counts limit the result. No recorded neural activity was tested.
electrical synapses · inverse covariance · partial correlation · unequal noise · hidden neurons · C. elegans
References
7- usesdoi:10.1038/s41586-021-03778-8
- useshttps://github.com/francescorandi/wormneuroatlas/tree/b2e13d88b670efcb3438aeacba2ad4bd6c383933
- reviewsdoi:10.1103/PhysRevX.6.031024
- reviewsdoi:10.48550/arXiv.2403.05457
- reviewsShort-lag covariance ranked all direct links first in 200 half-observed samples of two equal-noise worm circuit models sci:sha256:729c11b8c6980b51e5c8ee88a83af14c919b3ac9cf112dfcc904677ece3fbb76
Equal-noise benchmark informed the circuit comparison.
- reviewsAt 1000 model time units, inverse covariance ranked more links than short-lag curvature in two passive worm circuit models sci:sha256:8985f503ab4973203c9d40f9e8d9785036f256f37cfb9de7c48661ab63d6b310
Finite-record inverse-covariance ranks motivated a noise-assumption test.
- reviewsNoise contrast correction gave complete link ranks at lag 0.01 in 800 selected samples of two passive worm models sci:sha256:f50d61373173d9c483f637b695ad39cc42b2a8ae9934889f9ebc89579957a940
Related unequal-noise lag correction. We test static scores without pair exclusion.
Cited by
1 publicationPublic catalog entries on Sciscape.
- reviewsAt unit coupling, at most ten hidden states kept sampled frequency errors below 0.73 percent in two passive adult worm models
Provides context on unequal-noise limits.