From: Paul Victor Birke Subject: Re: Help : Simplex algorithm for non linear optimization Date: Tue, 05 Jan 1999 12:35:24 GMT Newsgroups: sci.math.num-analysis titou wrote: > > Hi > Could someone give me some information about the Simplex algorithm applied > to the resolution of minimisation of multi-variable non linear function. Good for only ~ 4 variables. See Viginia Torzson who fixed this algorithm in the early 1990s. Not recommmended. I found this out a long time ago in the early 1980s. You must get a hold of Viginia's agorithm from I think Rice University in Texas!! > Numerous authors prefered it to the Levenberg-Marquardt method. What are the > advantages and the limitations and where to find a good description of the > alogorithm ? Starts anywhere! Paul Birke, NN Researcher in Guelph ON CANADA > Thank you in advance. > Titou ============================================================================== From: mittelmann@asu.edu Subject: Re: Help : Simplex algorithm for non linear optimization Date: Tue, 05 Jan 1999 13:02:57 GMT Newsgroups: sci.math.num-analysis In article <76sh0t$2de$1@arcturus.ciril.fr>, "titou" wrote: > Hi > Could someone give me some information about the Simplex algorithm applied > to the resolution of minimisation of multi-variable non linear function. > Numerous authors prefered it to the Levenberg-Marquardt method. What are the > advantages and the limitations and where to find a good description of the > alogorithm ? > Thank you in advance. > Titou > > Hi, on http://plato.la.asu.edu/guide.html you find links to papers by M. Wright and M. Powell and a link to Powell's code COBYLA as well as to some others. Hans Mittelmann -----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own ============================================================================== From: "David Rothman" Subject: Re: Help : Simplex algorithm for non linear optimization Date: Tue, 05 Jan 1999 13:57:32 GMT Newsgroups: sci.math.num-analysis a good description of it can be found at: http://chem1.nrl.navy.mil/~shaffer/optsmpx.html numerical recipes also has a decent description of it, but the code is flawed so i wouldnt use it. the site above is instructive enuf to code it from there. its very easy. the advantages are that it is a non-derivative search function, and its conceptually very simple. the disadvantages are that it be require lots of iterations to converge (and thus run slowly) and it may not converge to the optimal pt. this is usually worked around by restarting the optimization at differing starting points. if u r clever with that, u can usually be effective. i use it or a very complex econometric problem. it works great...dave [quotation of original article deleted -- djr]