Curvature-Regularized Free-Energy Dynamics: A Unified Lyapunov Framework for Boundary-Coupled Systems
DOI:
https://doi.org/10.51094/jxiv.2082キーワード:
Free-energy principle、 Variational dynamics、 Curvature regularization、 Geometric regularization、 Curvature-constrained inference、 Energy landscape curvature、 Stability analysis、 Lyapunov dynamics、 Attractor geometry、 Nonlinear dynamical systems抄録
This paper develops a curvature-regularized formulation of free-energy dynamics that unifies inference-only and boundary-coupled systems within a single Lyapunov framework. We introduce a curvature field κ(x, y) that quantifies the local deformation of the free-energy landscape induced by external structural gradients. This curvature arises when internal inference interacts with boundary-coupled dynamics. The composite potential V(x, y) = F(x, y) + (λ/2)κ(x, y)^2 serves as a Lyapunov function guaranteeing V̇ ≤ 0 under standard regularity assumptions. The special case κ = 0 recovers pure inference, whereas κ > 0 describes boundary-regularized inference consistent with dynamic Bayesian stability.
利益相反に関する開示
The author declares no conflict of interest.ダウンロード *前日までの集計結果を表示します
引用文献
K. Friston. The free-energy principle: a unified brain theory? Nat. Rev. Neurosci.,11(2):127–138, 2010.
L. F. Barrett and W. K. Simmons. Interoceptive predictions in the brain. Nat. Rev.Neurosci., 16(7):419–429, 2015.
ダウンロード
公開済
投稿日時: 2025-12-04 10:46:15 UTC
公開日時: 2026-01-07 05:55:21 UTC
ライセンス
Copyright(c)2026
Sakai, Jun
この作品は、Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International Licenseの下でライセンスされています。
