Parameter Estimations via Genetic Algorithm in Multiresponse Nonlinear Models

Abstract

Multi-response non-linear models have been used for modelling functional relationship between dependent and independent variable(s) in most of applications. Parameters of multi-response non-linear models can be estimated by least squares (LS) method. Gauss-Newton, Levenberg-Marquardt and Steepest Descent are most widely used algorithms in LS method. These algorithms requires the condition that the function of independent variables can be differentiable at least two times. Also these algorithms have a risk of unreachable solution which depends according to the chosen starting point. In this study as an alternative to these algorithms, genetic algorithm have recommended for parameter estimation in multi-response non-linear models. And then the parameter estimation results of multi-response nonlinear model that obtained by least squares method and genetic algorithm were compared.