A new extension of Piyavskii's method to Hölder functions of several variables (Q2479217)

The following global optimization problem is considered: \[ \text{minimize }f(x)\text{ subject to }x\in P, \] where \(x\in\mathbb{R}^n\), \(f(x)\) is a real Hölder function with constant \(h> 0\) and exponent \(1/\beta\), \(\beta> 1\), and \(P\) is the Cartesian product of \(n\) one-dimensional closed intervals. The authors propose a new extension of \textit{S. A. Piyavskii}'s algorithm [U.S.S.R. Comput. Math. Math. Phys. 12 (1972), No. 4, 57--67 (1973; Zbl 0282.65052); An algorithm for finding the absolute minimum of functions. Theory of Optimal Solutions (Seminar, Kiev, 1967), No. 2, 13--24, Akad. Nauk Ukrain. SSR, Kiev (1967)] for solving the problem based on the construction of a piecewise linear sub-estimator. Convergence properties of the proposed algorithm are studied. The algorithm is demonstrated on a small numerical example.