A numerical proof of the Grunbaum conjecture

TL;DR

This paper provides a simplified numerical proof of the Grunbaum conjecture, which concerns bounds on projections in normed spaces, building on previous partial proofs and numerical studies.

Contribution

The paper offers a new, simpler proof of the Grunbaum conjecture, combining numerical analysis and previous theoretical work.

Findings

Confirmed Grunbaum conjecture with numerical methods

Provided a more accessible proof approach

Validated bounds on projections in normed spaces

Abstract

The Hahn-Banach theorem states that onto each line in every normed space, there is a unitary projection, and Kadec and Snobar proved (using John's ellipsoid) that onto each $n$-dimensional subspace of any real normed space, there is a projection with norm at most $\lambda_n \leq \sqrt{n}$. Grunbaum conjectured that $\lambda_2=4/3<\sqrt{2}$ and several attempts have been made to prove this conjecture: Konig and Tomczak-Jaegermann published a proof that was shown incomplete by Chalmers and Lewicki, who gave their own (a bit intricate) proof. Here is a simpler proof, mostly based on their works, and partially on a few numerical studies of extrema of functions of 3 variables.