Spectral Concentration at the Edge of Stability: Information Geometry of Kernel Associative Memory
High-capacity kernel Hopfield networks exhibit a \textit{Ridge of Optimization} characterized by extreme stability. While previously linked to \textit{Spectral Concentration}, its origin remains elusive. Here, we analyze the network dynamics on a statistical manifold, revealing that the Ridge corresponds to the Edge of Stability, a critical boundary where the Fisher Information Matrix becomes singular. We demonstrate that the apparent Euclidean force antagonism is a manifestation of \textit{Dual Equilibrium} in the Riemannian space. This unifies learning dynamics and capacity via the Minimum Description Length principle, offering a geometric theory of self-organized criticality.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
A Unified Matrix-Spectral Framework for Stability and Interpretability in Deep Learning
We develop a unified matrix-spectral framework for analyzing stability and interpretability in deep neural networks. Representing networks as data-dependent products of linear operators reveals spectral quantities govern…
Spectral-transport stability and benign overfitting for minimum norm interpolation
Benign overfitting describes the ability of minimum norm interpolating estimators to generalize despite fitting noisy data exactly. Existing characterizations depend on delicate spectral functionals of the population cov…
High-Dimensional Concentration and Retrieval Instability in Embedding Spaces: Implications for Retrieval-Augmented Generation
Embedding-based retrieval systems rely on the assumption that geometric proximity in highdimensional representation spaces reflects semantic relevance. However, high-dimensional geometry induces concentration phenomena t…
Stability and Concentration in Nonlinear Inverse Problems with Block-Structured Parameters: Lipschitz Geometry, Identifiability, and an Application to Gaussian Splatting
We develop an operator-theoretic framework for stability and statistical concentration in nonlinear inverse problems with block-structured parameters. Under a unified set of assumptions combining blockwise Lipschitz geom…
Random Spiking Neural Networks are Stable and Spectrally Simple
Spiking neural networks (SNNs) are a promising paradigm for energy-efficient computation, yet their theoretical foundations-especially regarding stability and robustness-remain limited compared to artificial neural netwo…