| The purpose of this thesis is to examine different methods of curve-fitting through the process of nonlinear least squares. Specifically, Newton's method, Gauss Newton's method, Levenberg-Marquardt, Quasi Newton methods, and Nonlinear Conjugate Gradient method. The differing convergence rates, as well as necessary initial conditions are explored for a variety of methods. The concepts of the line search and trust region are also examined. |
