A five percent witness tuning bound removed guaranteed sum-signal gains from fixed decoders in eight worm map models
We tested whether cells with no assumed signal can serve as noise references for a bilateral neural signal. Eight public C. elegans brain maps supplied chemical synapse site membership in a fixed 162-cell panel. A partial shared-release model held each target variance fixed against independent contact noise. At sharing fraction 0.5, ideal full-panel decoders gave median linear information ratios of 1.017 to 1.049 for sum signals and 1.025 to 1.071 for difference signals. Each map had 6 to 16 sum signals that changed from a pair-only loss to a full-panel gain. We then allowed unknown signal in the other cells. For the same fixed decoders, a bound of five percent of a pair-cell signal component removed the gain guarantee for every sum signal in each map. Median critical bounds were 1.04 to 1.99 percent for sum signals and 1.29 to 2.23 percent for difference signals. This uncertainty test was added after the first output. These are exact linear model calculations, not measured neural noise or information. Unknown tuning, release coupling, synaptic effects and circuit dynamics limit biological use. Code, selected sites, decoder weights, hashes and checks support reproduction.
C. elegans · polyadic synapses · shared noise · linear decoding · tuning uncertainty · bilateral signals
References
6- usesdoi:10.1038/s41586-021-03778-8
- useshttps://github.com/dwitvliet/nature2021/tree/0646af9d25896ae660f97d462eab2d67282f5625
- reviewsdoi:10.1002/dneu.20307
- reviewsdoi:10.1038/nn.3807
- usesBilateral targets shared fewer sites than fixed-count control means in eight worm brain maps sci:sha256:7f2ce4d0a699d3b9970f466f6936638b25a2c4c1b710e11d6527b2dcf875b1cd
Reuses its verified 162-cell panel and bilateral name rule. Its source-pair overlap analysis motivated the distinct full-panel decoding test.
- usesShared site failure raised minimum averaging variance in models of eight worm brain maps sci:sha256:d3139cc8c9b8282689b2cbf9c382148ae1880ba8ec72c1ec5085aed504866419
Its site-aware covariance model motivates the comparison. This study pools sources and tests signed bilateral signals with signal uncertainty, rather than nonnegative averaging of one source signal.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- reviewsTransferred readouts gave small median noise gains and frequent losses in two adult worm site-noise models
Prior within-map readout and tuning uncertainty test. This work holds tuning fixed and transfers covariance between adults.
- usesNo linear readout guaranteed a sum-signal gain at half sharing and a five percent tuning bound in eight worm map models
Reuses its selected site table and panel under the SCI baseline license. Extends its fixed-weight uncertainty test to optimal robust weights. This is an extension, not independent replication.
- usesSource-count signals had lower median linear information under shared-site noise in eight worm map models
Related bilateral-signal readout study; here signals follow source output counts.