SGI · SIG-NmNN · Research
The breathing neuron. Activation regions that grow with their input.
SIG-NmNN — Spike-Intensity Growth Non-Monotonic Neural Networks: a framework where neuronal activation regions dynamically expand based on accumulated spike intensity. Neurons breathe with their input — contracting during low activity, expanding during high activity — and naturally produce fractal structures through power-law scaling when nested hierarchically.
Experiment 1 · The breathing neuron
The blue torus is the activation volume; the sphere is the input. Watch the torus expand as intensity builds. The input glows green when inside the volume and red when outside.
Experiment 2 · Stability dynamics
Comparison of integration dynamics with discrete time step Δt = 0.1 (stable) vs Δt = 2.1 (unstable), under constant input 1.0.
Stable (Δt = 0.1)
Unstable (Δt = 2.1)