Spectral Selection under Exponential Tilting: Covariance Dynamics and Edge Asymptotics

Fuente: Zenodo
Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Morissette, Louis
Format: Recurso digital
Sprache:Englisch
Veröffentlicht: Zenodo 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866902337761574912
author Morissette, Louis
author_facet Morissette, Louis
contents <p>We study the dynamical structure induced by exponential tilting of a positive measure μ on [0,∞) with finite second moment. The tilted family ν_t(dλ) ∝ e^{−λt} dμ(λ) satisfies the exact covariance identity<br><br><span>d/dt E_t[g] = −Cov_t(λ,g),</span><br><br>from which the variance dissipation law<br><br><span>r′(t) = −Var_t(λ) ≤ 0</span><br><br>follows, where r(t) = E_t[λ].<br><br>The family (ν_t) is a one-parameter exponential family whose Fisher information satisfies<br><br><span>I(t) = Var_t(λ) = −r′(t).</span><br><br>This identity provides a closed scalar law governing spectral selection dynamics.<br><br>We prove convergence r(t) → λ* = inf supp(μ) and classify rates: exponential under a spectral gap, and algebraic r(t) − λ* ~ β/t under regular variation μ([λ*,λ*+x]) ~ x^β L(x). We establish the global dissipation identity<br><br><span>∫₀^∞ Var_t(λ) dt = r(0) − λ*,</span><br><br>under a finite first-moment condition. For bi-atomic spectra, the dynamics closes exactly as a Riccati equation r′(t) = −(r−p)(q−r), saturating the variance bound.<br><br>To our knowledge, the explicit formulation r′(t) = −I(t) as a standalone identity governing spectral selection does not appear explicitly in the literature.<br><br>This framework provides a minimal and exact backbone for spectral flow, linking covariance dynamics, Fisher information, and asymptotic edge selection.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_19074037
institution Zenodo
language eng
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle Spectral Selection under Exponential Tilting: Covariance Dynamics and Edge Asymptotics
Morissette, Louis
Laplace mixtures
replicator equation
hazard rate dynamics
Riccati equation
variance dissipation
spectral dynamics
covariance dynamics
completely monotone functions
replicator dynamics
Price equation
moment cascade
<p>We study the dynamical structure induced by exponential tilting of a positive measure μ on [0,∞) with finite second moment. The tilted family ν_t(dλ) ∝ e^{−λt} dμ(λ) satisfies the exact covariance identity<br><br><span>d/dt E_t[g] = −Cov_t(λ,g),</span><br><br>from which the variance dissipation law<br><br><span>r′(t) = −Var_t(λ) ≤ 0</span><br><br>follows, where r(t) = E_t[λ].<br><br>The family (ν_t) is a one-parameter exponential family whose Fisher information satisfies<br><br><span>I(t) = Var_t(λ) = −r′(t).</span><br><br>This identity provides a closed scalar law governing spectral selection dynamics.<br><br>We prove convergence r(t) → λ* = inf supp(μ) and classify rates: exponential under a spectral gap, and algebraic r(t) − λ* ~ β/t under regular variation μ([λ*,λ*+x]) ~ x^β L(x). We establish the global dissipation identity<br><br><span>∫₀^∞ Var_t(λ) dt = r(0) − λ*,</span><br><br>under a finite first-moment condition. For bi-atomic spectra, the dynamics closes exactly as a Riccati equation r′(t) = −(r−p)(q−r), saturating the variance bound.<br><br>To our knowledge, the explicit formulation r′(t) = −I(t) as a standalone identity governing spectral selection does not appear explicitly in the literature.<br><br>This framework provides a minimal and exact backbone for spectral flow, linking covariance dynamics, Fisher information, and asymptotic edge selection.</p>
title Spectral Selection under Exponential Tilting: Covariance Dynamics and Edge Asymptotics
topic Laplace mixtures
replicator equation
hazard rate dynamics
Riccati equation
variance dissipation
spectral dynamics
covariance dynamics
completely monotone functions
replicator dynamics
Price equation
moment cascade
url https://doi.org/10.5281/zenodo.19074037