Vanishing Hermite Polynomials in Global Optimization

George A. Vazmin

Abstract: Weighted Hermite polynomials row for function symmetrized around objective (unknown) point opens a way for a localization of the global maximum position. This approach creates subsidiary function guiding to the most prominent extrema (up to 3). The guidance function is generalized to multi-dimensional space. Several analytical tools (suppression of noise, space constriction, etc) are described in this technology.

Keywords: Hermite polynomials, symmetrization, guidance, multi-dimensional, noise suppression, scaling, space collapse, recursive iterations, poor statistics.

Pages: 33 – 44 | Full PDF Paper