At gain 0.85 and one leak time, added sensory ascending input allowed norm growth in a full FlyWire v783 count model

We tested the effect of allowed initial cells on norm growth in one FlyWire brain model. The model kept all 139,255 proofread neurons and 54,492,922 chemical contacts from release 783.0. Each target row was divided by its full input count. All states had equal leak. Coupling gain was 0.85. The primary time was one leak time. The exact sensory class had 16,904 cells. For initial activity in this class, every norm ratio was below 0.897202. This bound permits any fixed edge signs at the stated count magnitudes. The largest ratio found with the inherited signs was 0.895486. Adding initial access to 612 sensory ascending cells gave a saved ratio of 1.221739. Adding 1,750 ascending cells gave 1.511497. Every cell remained as a relay and in the output. The first expanded pattern placed 99.78% of squared output norm in sensory ascending cells. The allowed input class thus changed the growth result. The bound for the exact sensory class does not extend to all sensory pathways. Positive matrix bounds, guarded arithmetic, dense checks and an independent exponential check support these results. These are conditional model responses at fixed gain and time. They are not measured physiological gain. The package preserves code, source hashes, metadata, response vectors, figures and repeat instructions.

connectome · FlyWire · sensory neurons · input access · transient growth · positive matrix bound · reproducible benchmark