Why Learned Recurrent Inference Dynamics Resist a Transfer Operator Bridge to the Riemann Hypothesis: An Honest Assessment

Abstract

This essay examines the recurrent dynamical system and spectral Jacobian structure described by Yun et al.

Overview

This essay mines the neuroscience and machine learning literature for structural patterns that might suggest approaches to the Riemann Hypothesis (RH). The source paper proposes a sparse coding model of visual cortex that learns a non-factorial prior via denoising score matching. The resulting recurrent dynamical system admits a spectral decomposition of its Jacobian, which the authors use to explain "natural structural deformations."

The Attempted Bridge

We investigate whether the recurrent dynamics and the associated spectral operator (the Jacobian of the denoiser) can be mapped onto the transfer operators and dynamical zeta functions used in the spectral approach to RH. Ruelle’s theory connects the zeros of dynamical zeta functions to the spectrum of transfer operators associated with expanding maps. The superficial similarity lies in the shared language of "spectra," "recurrent dynamics," and "deformations."

The Assessment

The analogy is rated a SUGGESTIVE METAPHOR at best, but effectively fails to reach the level of a formal analogy. The source dynamics are gradient-like inference flows that converge to fixed points; they lack the chaotic mixing, periodic orbits, and expanding properties required for Ruelle-Selberg zeta functions. The Jacobian is a local, data-dependent self-adjoint matrix with real eigenvalues, whereas the RH requires a global operator with complex eigenvalues (or zeros) on the critical line.

Key Finding

The critical missing ingredient is an Euler product structure or a sum over periodic orbits. Without these, the spectral data from the learned model cannot be organized into a zeta function whose zeros encode arithmetic information. This honest negative outcome clarifies the boundary between dissipative inference and chaotic dynamics.

This essay was produced by an automated research pipeline and has not been peer reviewed; conjectures herein are unproven.