ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let k∈Nk\in\mathbb{N}. For each d∈Nd\in\mathbb{N}, let U=diag⁡(1,ζd,ζd2,…,ζdd−1)U=\operatorname{diag}(1,\zeta_d,\zeta_d^2,\dots,\zeta_d^{d-1}), where ζd=exp⁡(2πi/d)\zeta_d=\exp(2\pi \mathbf{i}/d) is a primitive root of unity of order dd. For 0<ρ<10<\rho<1, consider the linear system from the source with matrix UU and initial value x0=0x_0=0, and let rn(k)r_n^{(k)} denote the residuals produced by Orthomin(k)(k). Sharpness conjecture. For every k∈Nk\in\mathbb{N}, there exist dk∈Nd_k\in\mathbb{N} and 0<ρk<10<\rho_k<1 such that, for every ρ∈(0,ρk)\rho\in(0,\rho_k) and every d≥dkd\ge d_k,

lim⁡n→∞∥rn+1(k)∥∥rn(k)∥=ρ.\lim_{n\to\infty}\frac{\lVert r_{n+1}^{(k)}\rVert}{\lVert r_n^{(k)}\rVert}=\rho.

The result is proved in the paper for k=1k=1, while for k≥2k\geq 2 the authors report strong numerical evidence. If true, it shows that Orthomin(k)(k) need not converge asymptotically faster than Orthomin(1)(1) in general.

Record

Comments

Nothing has been published against this problem yet, and nothing has been said about it.

Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.

Sign in

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.