NOMAD, C++ software for derivative-free optimization

The underlying algorithm in NOMAD is MADS, a direct search algorithm with a very strong theoretical and practical justification. The user provides �blackbox� code that NOMAD can call to evaluate the objective and constraints for the nonlinear programming problem the user wishes to solve. Convergence is guaranteed under very general hypotheses.

Advantages NOMAD does not need sensitivity information from the user. This is a major advantage over methods that are more efficient that NOMAD if such information is known. Other algorithms that require the same user supplied information as NOMAD are the Nelder-Mead method, which is very effective, but can fail in theory and practice, and the quite ineffective genetic or evolutionary algorithms. The novelty is the use of MADS as the underlying algorithm.

Potential Applications . Low dimensional, nonsmooth optimization problems in engineering design

Inventor(s): John E. Dennis

Type of Offer: Licensing



« More Software Patents

---