Simulated annealing optimization in nonlinear regression: Algorithm and software (Q2726326)

The adjustment of nonlinear regression equations is studied. A grid of the parameter space should be constructed. Once the discretization is defined a search procedure may be used as an alternative method for estimating the regression parameters. The authors propose to use a simulated annealing procedure. The corresponding algorithm is presented and its performance is illustrated using well known data sets obtained from \textit{A. Antoniadis, J. Berruyer} and \textit{R. Carmona} [Régression non linéare et applications. Paris, Economica (1992)], \textit{D. Battes} and \textit{D. Watts} [Nonlinear regression analysis and its applications, Wiley (1988; Zbl 0728.62062)] and \textit{R. Tomassone}; \textit{S. Audrain}; \textit{E. Lesquoy-de Turckheim}; \textit{C. Millier} [La Régression: Noveaux regards sur une ancienne méthode statistique. (1992; Zbl 0788.62058)].