Hadamard growth problem: linear growth factor

Establish whether the Gaussian elimination growth factor ρ(H_n) equals n for every Hadamard matrix H_n. Determine if the maximal intermediate-element growth under Gaussian elimination (as defined in the paper) is exactly linear in n across the class of Hadamard matrices.

Background

The growth factor ρ(A) controls numerical stability in Gaussian elimination. For Hadamard matrices, it is widely believed that the worst-case growth equals n. This has been proved for all Sylvester Hadamard matrices, and for general Hadamard matrices only up to n = 16.

The paper studies butterfly matrices and introduces butterfly Hadamard matrices, providing new structural insights and exact growth computations in specific settings, but the general Hadamard growth problem remains open.

References

A similarly famous open problem for numerical analysis involves the growth problem for Hadamard matrices, where it is believed $\rho(H_n) = n$. This has been established for all Sylvester Hadamard matrices (see ), while for general Hadamard matrices this has only been proven up to $n = 16$ . The growth problem for Hadamard matrices is a sub-question for the growth problem when restricted to orthogonal matrices. For instance, the orthogonal growth problem remains open in GEPP even, which was recently explored in .

Complete pivoting growth of butterfly matrices and butterfly Hadamard matrices (2410.06477 - Peca-Medlin, 9 Oct 2024) in Section 1 (Introduction)