Vector Space of Cycles

A new variational framework treats cyclic interactions as elements of a Hilbert space, allowing researchers to project, average, and compare recurrent organization in high-dimensional systems, according to a paper submitted on 6 June 2026 [1]. Most statistical and machine learning methods for directed interactions focus on pairwise effects among variables, and existing cyclic models represent feedback primarily through node-level dependencies [1]. The authors argue this makes large-scale recurrent organization difficult to estimate and compare, a limitation they describe as particularly acute in biological and neural systems where interactions are highly recurrent and involve many overlapping cycles [1]. The framework represents directed interactions as edge flows on a simplicial complex and evolves them under an energy-minimizing dynamical system [1]. The resulting dynamics separate transient interaction components from persistent harmonic flows, yielding a low-dimensional cycle space that captures stable recurrent organization [1]. In graph theory, the cycle space of an undirected graph is the set of its even-degree spanning subgraphs, describable algebraically as a vector space over the two-element field, with its dimension equal to the circuit rank of the graph [5]. The new work extends such concepts to directed, recurrent settings. Rather than enumerating individual cycles, the proposed framework represents cyclic interactions as elements of a Hilbert space, enabling projection, averaging, comparison, and population-level statistical inference [1]. The authors establish theoretical properties of the harmonic projection, including characterization of the cycle space, variance reduction, and population inference [1]. Simulations demonstrate substantially improved recovery of cyclic structure in dense recurrent systems compared with existing directed-interaction methods [1]. Applied to resting-state fMRI from 400 human subjects, the framework reveals reproducible large-scale cyclic organization that is not detectable through edgewise averaging [1]. The paper was posted on arXiv, an open-access repository that hosts preprints in mathematics, physics, computer science, and related fields and that, as of November 2024, received about 24,000 submissions per month [10]. The authors state the results provide a scalable statistical framework for studying recurrent interactions in high-dimensional dynamical systems [1].