Last modified: 2007 Jan 15

Optimization Routines

Don't overlook the routines in the Run-Time Library that comes with each copy of GAUSS. QNewton() is a function implementing the BFGS method for optimization of an arbitrary function. And there's also an NLP program called SQPSolve() for optimization of functions with general linear/nonlinear, equality/inequality constraints. (Schoenberg recommends using SQPSolve() even when the problem is unconstrained.) These two both have an 'mt' version: Qnewtonmt, and Sqpsolvemt. Qnewtonmt and Sqpsolvemt use the relatively new PV and DS structures. The PV structure handles parameters, and the DS structure handles data. Aptech provides details.

• Sandeep Mohapatra's programs for Data Envelopment Analysis developed for his (co-authored) paper Dynamic Productivity, Efficiency and Technical Innovation in Education: A Mathematical Programming Approach Using DEA.
  Comment: requires the Constrained Optimization package.
  Comment: the DEA programs are currently in a somewhat odd form. They are zipped together as a collection of HTML files, which display formatted code.
• Nelder-Mead Polytope program in a .zip file (Schoenberg May99)
• Nelder-Mead Simplex search, generally following Numerical Recipes. (McCulloch May99)
• One Hidden Layer Feedforward Neural Networks (Zeng Nov98)
• Simulating Neural Networks (Ho Jul00)
• Simulated Annealing by E.G.Tsionas (Sep95) includes two programs: the simulated annealing routine and an example with a bimodal nonlinear likelihood. (Also see Bill Goffe's original Fortran version.)
• GAUSS versions of the optimization routines in Numerical Recipes, written by Bo Honore and Ekaterini Kyriazidou, corrected 9/9/96. Also check Bo Honore's site
• Davidson, Fletcher, Powell minimization routine essentially as transcribed by Honore & Kyriazidou but slightly modified to increase ease of use by Isaac (Aug96).
• Neural Net "Backpropagation" (Roncalli Nov96).
• Quadratic Programming (Dittmar, Sep96)
• Finding roots of a function (van der Ende, Sep96)
• Finding zeros of a function (Isaac Dec03)