CGO

Many problems in the sciences, engineering disciplines, economics and planning, medicine and health, and the communication industry can be expressed as global optimization problems. Indeed, the global optimization problem is often the ‘driver’, i.e., the overarching problem to be solved. For example, maximize profit, minimize risk, find the shortest path, determine the energy-minimizing ‘fold’ for a molecule, determine the ‘best’ design for a craft, subject to material and cost constraints, and so on. ‘Global’ indicates that the best solution is sought over a pre-defined region.

The Cayuga Global Optimizer (CGO) is a powerful tool, combining several technologies, to search for global optimizers for continuous optimization problems. CGO is designed for use in the MATLAB environment and has several outstanding features:

· Derivatives can be computed automatically using the Cayuga tool ADMAT: there is no need to supply derivative functions (or derivative approximation schemes).

· Phase 1 improves on a set of user-supplied starting points (or randomly-generated starting points) using one of several simulated-annealing methodologies (some variants require function evaluations only) and approximation schemes.

· Phase 2 employs Matlab local minimizers starting from the endpoints in Phase 1, to obtain a set of accurate local minimizers.

· Both Phase 1 and Phase 2 allow for easy use of parallelism: see Cayuga Research package PCGO.

· CGO allows for significant flexibility: the user can tailor a global optimization approach by using one of several options for Phase 1 (or replace the entire Phase 1) with a user-supplied procedure for improving on starting points; there is MATLAB-flexibility in choosing or tailoring the local minimization procedures in Phase 2.

· User supplied functions can be MATLAB M-files or C, Fortran functions wrapped as MATLAB Mex-files.

· Either Phase 1 or Phase 2 of CGO can be employed independently.

· CGO is driven either using an easy-to-use graphical user interface (GUI), or Matlab function calls.