High frequency reduced hidden-path bias but raised noise error in passive models of two adult worm maps

We tested whether frequency-dependent port responses can separate direct electrical links from hidden paths in passive circuit models. A Schur-complement proof gives quadratic real-inverse bias decay for strictly positive hidden capacitance. A two-frequency estimate cancels the leading bias term. We tested 12 half-observed masks in each of two partial adult worm maps under uniform and variable capacitance. For uniform capacitance, exact DC median weight errors were 19.11 and 22.86 percent. At 30 times the full-network maximum relaxation rate, single-frequency errors fell to 0.00619 and 0.00630 percent. An absolute complex impedance noise floor reversed the high-frequency benefit. At relative noise 0.0001 and frequency ratio one, two-frequency errors were 2.46 and 5.23 percent, compared with 4.08 and 4.16 percent from two trials at one frequency. Thus extrapolation did not give a consistent gain. All positive-noise settings favored a base frequency below the highest tested value. This is a conditional method benchmark with assumed conductances and noise, not measured neural responses or a new general identification theorem. Code, aggregate results, proof, checks, source hashes and retrieval instructions are included; external records remain outside the bundle.

electrical coupling · hidden nodes · passive circuits · frequency response · connectome · noise · C. elegans