Two leak levels reduced exact-response weight error in two passive worm models under a uniform leak change

We tested a two-leak estimate of direct electrical weights in passive models of two partial adult worm maps. A Schur-complement calculation shows that a known uniform leak change reduces full-matrix hidden-path bias in positive-semidefinite order. It does not guarantee better estimates at each pair. In 30 half-observed masks per map at base leak one, exact-response median weight errors fell from 13.92 and 14.78 percent at double leak to 8.28 and 9.43 percent with the two-leak estimate. Some pair errors grew and some estimated weights were negative. Under an assumed absolute response noise floor of 0.01 times the low-leak RMS response, the two-leak estimate lost to repeated high-leak measurements in all 60 masks. A change restricted to observed cells provides no added direct-weight information. The method requires calibrated full response matrices and a known leak change in hidden cells that preserves coupling. This is a conditional numerical method test, not measured neural conductance or activity. Original code, proof, aggregate evidence, source hashes and retrieval steps support reproduction. External records remain outside the bundle.

electrical coupling · hidden neurons · leak conductance · Kron reduction · connectome · passive circuit · C. elegans