|
Abstract:
|
For many business applications, it is commonly the case that derivative information for any objective function of interest is not available. In the absence of derivative information, one must rely on optimization algorithms which only require that function evaluations be available. These algorithms are known as direct search methods. We implement the direct search algorithm of Torczon (1989, 1991) in S-plus and explore its utility in several interesting applications. Torczon has proven the superiority of her algorithm to the widely referenced Nelder Mead algorithm. Among many interesting features of the Torczon algorithm, it lends itself well to problems where the constraints are quite complicated.
|
