Work

Eigengap Stability and Residual Certificates for Least-Squares Lie Sector Approximation in the Symmetrized Sedenion Khanhnion Model

Year: 2026
Type: preprint
Abstract: We study the uniqueness and perturbation stability of the least-squares maximal Lie sector selected by two principal eigenvector lines in the standard-real two-channel symmetrized sedenion model. A positive eigengap in each channel ensures a unique fitted sector. Under symmetric operator-norm perturbations smaller than half these gaps, we prove explicit line-angle and sector-projector bounds, together with quadratic excess-loss estimates. We also give a posteriori gap certificates from an observed covariance matrix and residual-based certificates for a candidate direction. A repeated-eigenvalue example shows that arbitrarily small changes can select widely separated minimizing lines, even when the optimal value remains stable. The estimates are elementary rank-one Davis-Kahan-type bounds and are explicitly placed within established spectral perturbation theory. No priority or optimal-constant claim is made.
Language: English
Open Access status: green
OpenAlex ID: W7236703539
Cites:
Cited by:
Related to: 10