A second inhibitory input reduced model residual by at least 0.05 in one of 11 selected Allen slice panels

We tested ideal linear cancellation at targets with one excitatory and two inhibitory inputs in Allen Synaptic Physiology r2.1 slice data. Eleven target-and-voltage panels used 33 passed current-clamp fits from 10 targets and nine inferred donors. With finite rise and measured onsets, separate nonnegative inhibitory gains reduced mean residual energy from 0.20520 to 0.19790 of excitatory-input energy. Only one panel had a reduction of at least 0.05. Seven panels used only one input at the optimum. A common gain that kept the measured inhibitory amplitude ratio gave mean residual 0.36315 and was worse than the best single input in nine panels. An overlap criterion explains when a second input helps. Independent sampled-kernel least squares agreed within 8.12e-09. These are conditional model calculations, not measured cancellation. Sparse selected slice circuits, fit error, separate stimulation, and linear voltage addition prevent population or intact-brain claims. The two human panels share one target and donor. Code, extracted records, hashes, source links, checks, and a figure support reproduction.

Allen Institute · convergent inhibition · response shape · nonnegative gain · cancellation bound