A single sensor can be blocked, jammed or simply unlucky. Here four SPNM nodes watch for the same rhythm and share accumulated evidence with their neighbours. Block one node, vary the cooperation strength, and explore what distributed SPNM dynamics gain — and what they risk.
Network-level dynamics. Each node is an SPNM resonant unit; Ei is its accumulated evidence, which leaks over time and is coupled to neighbouring nodes through a bounded tanh term. This is a distributed evidence layer built around the SPNM oscillator, not a separate fundamental neuron formulation.
| N1 | N2 | N3 | N4 | |
|---|---|---|---|---|
| Alone: time to alert | ||||
| Cooperating: time to alert |
A simulation of the equations above, not a hardware measurement or a benchmark. Each node is the resonant unit of the earlier demo (ż = (−a + iω₀)z − β|z|²z + γu, a = 0.6, β = 1.2) driven by its own input: a 1 Hz rhythm plus noise, or arrhythmic clutter of similar power. A blocked node receives 10% of the rhythm. Evidence E accumulates the rhythmic energy of the node and leaks at rate λ = 0.3; a node raises its alert when E reaches 0.55 and releases it below 0.30. The coupling term uses tanh(E) of the neighbours, a bounded choice made for this illustration; with it, E stays bounded for any σ, and σ·dmax < λ is the usual sufficient condition for a guaranteed stable evidence level, marked on the slider. Time is measured from the start of the target window (3 s). Parameters are illustrative, not fitted to any dataset, and the noise is fixed per seed.