The q-Exponential Machine

A decaying curve doesn't have to choose between "exponential" and "power-law." It can be neither, on its own β€” it can be a mixture of exponentials: countless simple decays e^(-λt), each running at its own rate λ, summed together with weights given by a spectral density ρ(λ). When ρ(λ) is a Gamma distribution with shape parameter k, the sum works out in closed form to a Tsallis q-exponential β€” and in the paper's 1D absorbing-boundary geometry, exactly q = 5/3 at a specific k. Drag k below and watch the spectrum and the resulting decay curve morph live, computed by literally summing the mixture on a grid β€” no lookup tables.

Based on: "A spectral model of power-law decay in natural and engineered systems" β€” arXiv:2606.08342v1

k = 1.00  |  q = 1+1/k = 2.00

Panel 1 — Spectral density ρ(λ)

Gamma(k, θ=1/k) probability density over decay rate λ. Mean rate fixed at 1 so only the shape changes as k moves.
probability mass per bin spread (CV = 1/√k): —

Panel 2 — Summed decay curve N(t), log–log

N(t) = ∫ ρ(λ) e-λt dλ, computed by numerical integration over the same spectrum, every frame.
your mixture N(t) pure exponential e-t paper's q=5/3 curve
πŸ€
scenario
—
Illustrative toy grounded in the paper's real spectral-mixture math — not fitted to actual NBA data. Any resemblance to a specific player's shooting slump is for flavor only.