Work

Least-Squares Lie Sector Approximation and Quantitative Jacobi Tests in the Symmetrized Sedenion Khanhnion Model

Year: 2026
Type: preprint
Abstract: We seek one maximal rotation-invariant Lie sector approximating a finite family of such sectors in the standard-real two-channel symmetrized sedenion model. Under the explicitly chosen Euclidean metric and mean squared Frobenius projector loss, the optimization separates into two classical rank-one projection-mean problems solved by top eigenvectors. We give the exact optimum, characterize nonuniqueness, and bound the change in its value under feasible signed-Gram perturbations without a same-rank assumption. We then define sampled three-element Jacobi tests whose squared average controls this optimal approximation error within explicit constants. In contrast, the supremum of the normalized three-element test over the entire generated algebra jumps from zero to four as soon as either internal imaginary span has dimension at least two. The note therefore separates small sampled residuals, good single-sector approximation, and exact Jacobi closure. Projector means and eigenvector optimization are established tools; no novelty, optimality beyond the specified loss, or computational advantage is claimed. 1. Model, metric, and relationship to Paper 25 Let E = ℝ¹⁵, J = ℝ ⊕ E with (a,x) ∘ (b,y) = (ab − ⟨x,y⟩, ay + bx), W = J × J, and L = V ⊗ W for V = ℝ³. Its bracket is [v ⊗ a,w ⊗ b] = 2(v × w) ⊗ (a ∘ b). A line ℓ defines Cℓ = ℝ1 ⊕ ℓ and a maximal invariant Lie sector Mℓ,m = V ⊗ (Cℓ × Cm). We use the standard positive Euclidean vector-space inner products on J, W, and L, not a positive-form replacement of the algebra multiplication. Given n ≥ 1 sectors Mᵢ = Mℓᵢ,mᵢ, including possible repetitions, we minimize their mean squared projector distance to one sector Mℓ,m. Paper 25 addresses conditional reconstruction and rank instability (Nguyễn Kim, 2026). Here the objective is geometric approximation, not exact recovery of an original rank or bracket-preserving repair. Projection-distance subspace means solved by principal eigenvectors are established in the literature; Besson, Dobigeon and Tourneret (2011) give this principle in a Bayesian subspace-estimation setting. We prove the finite deterministic specialization directly. Figure 1 is on the final page.
Language: English
Open Access status: green
OpenAlex ID: W7236688729
Cites:
Cited by:
Related to: 10