Rate-Function Structure of the Correlated Ensemble Solver: Global Structural Functional, Spectral Sensitivity, and Dominant-Mode Dominance
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Two structural probes are applied to the Correlated Ensemble Solver (CES) to characterise the mathematical object it computes. The first probe tests block-composition factorisation: does log P(A+B) equal log P(A) + log P(B) when A and B are exactly uncoupled? The second tests spectral representation invariance: does P depend only on the eigenvalue spectrum of the correlation matrix, or on its full topology? The results are definitive on both questions. CES outputs do not factorise across independent subsystems: the maximum deviation reaches 0.556 on exactly block-diagonal matrices with zero cross-block entries. Since standard joint exceedance probability must factorise over uncoupled blocks by ordinary probability theory, this result establishes that the CES computes a global structural functional—a quantity sensitive to the assembled system as a whole rather than the product of subsystem probabilities. This is not a pathology but a feature: the intensive rate function decreases systematically with N (from 0.145 at N = 5 to 0.023 at N = 50), revealing sub-extensive scaling consistent with a solver that integrates over the full assembled correlation structure. Spectral invariance fails dramatically: relative probability deviation reaches 0.73 across topologies with identical eigenvalue spectra, with hub, chain, random, and equicorrelation wiring giving P values of 0.092, 0.170, 0.222, and 0.341 at the same threshold, scale parameter, and spectrum. Crucially, the sensitivity to sub-dominant eigenvalues at fixed dominant eigenvalue is negligible. The CES therefore exhibits dominant-mode dominance: it integrates over the dominant signal subspace of the assembled correlation structure, not a full combinatorial enumeration and not a simple subsystem product.
Fields of Study & Subjects
Statistical Mechanics and Entropy
Matrix Theory and Algorithms
Probability and Statistical Research
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 28 March 2026
- Language
- EN
- Document type
- Preprint