A hybrid global optimization algorithm is proposed aimed at the class of objective functions with properties typical of the problems of non-linear least squares regression. Three components of hybridization are considered: simplicial partition of the feasible region, indicating and excluding...