From: spellucci@mathematik.tu-darmstadt.de (Peter Spellucci) Subject: Re: Is there a Gauss-Newton method??? Date: 23 May 2000 10:54:50 GMT Newsgroups: sci.math.num-analysis Summary: [missing] In article , p_mclean@postoffice.utas.edu.au (Patrick McLean) writes: |> If a set of equations in several unknowns is linear in some variables and |> nonlinear in others. Is there a way to solve the equations which is like |> Gaussian elimination in the linear variables and like Newton's method in |> the nonlinear variables. |> what you require is known as separable least squares problem. there are several methodes for dealing with that, also ready to use methods. e.g. dqed does this. look up http://plato.la.asu.edu/guide.html in problems/software special codes for least squares problems. deqd is good only for small residual problems. for large resudual problems it is better to use a general solver, also listed there. hope his helps peter ============================================================================== From: Christof Pflumm Subject: Re: Is there a Gauss-Newton method??? Date: 23 May 2000 10:27:58 +0200 Newsgroups: sci.math.num-analysis p_mclean@postoffice.utas.edu.au (Patrick McLean) writes: > If a set of equations in several unknowns is linear in some > variables and nonlinear in others. Is there a way to solve the > equations which is like Gaussian elimination in the linear variables > and like Newton's method in the nonlinear variables. You could have a look at "Computer oriented algorithms for solving systems of simultaneous algebraic equations" by Kenneth M. Brown in "Numerical solution of systems of nonlinear algebraic equations (1973)", page 281 edited by George D. Byrne and Charles A. Hall. Bye, Christof