NOS Demo B: Event coupling vs coupling-through-nonlinearity

Two side-by-side simulations of the same NOS cell dynamics (v,u) with identical topology and noise. Left uses the paper-style event-based coupling: I(t)=k W S(t-τ). Right uses a deliberately deprecated continuous coupling: I(t)=k W g(v(t)) (a common conceptual mistake). This demo is forward simulation only, not a training surrogate.

Controls

36
6
Keeps topology comparable and avoids sign issues.
Separates topology effects from gain scaling.
k = 0.400
τ₀ = 6.0 ms
T = 2.0 s
Smaller dt is slower but more stable.
NOS parameters (subset)
These match the form used in nos.py:
dv = f_sat(v)+ (β−λ−χ)v − u + γ + χ v_rest + I,
du = a(bv−u) − μu, spike if v_new ≥ θ, then pullback reset to c_reset with u += du_spike.
0.10
0.00
0.80
0.40
0.00
0.02
1.00
1.00
0.20
1.00
6.0
0.00
0.50
0.45
0.02
ρ(W) (raw)
–
ρ(W) (used)
–
k ρ(W)
–
Event panel burst index
–
Deprecated panel burst index
–

Output

Top: raster plots (left event coupling, right deprecated coupling). Bottom: population spike count (binned).