Two algorithms for global optimization
Mathematical models are used in many fields to gain insight and perspective on systems. Systems performance is measured by an objective function. Optimization techniques generate values for the decision variables which maximize (or minimize) system performance. Global optimization is a branch of mathematical programming in which these decision variables are unconstrained. Examples of global optimization include minimizing a total cost function, optimal portfolio selection and facility location. Many global optimization problems have non-linear objective functions and may be neither convex nor concave. In particular , the objective function may have many local maxima. By partitioning the feasible region into regions of attraction, the application of a local optimization procedure once in each such region will provide all local maxima; and the largest of these local maxima is a global maximum. This research develops and investigates two algorithms for solving global optimization problems by partitioning the feasible region. One algorithm uses Cluster Analysis techniques in the framework of the Artificial Intelligence heuristic Tabu Search; the second algorithm uses mixtures of Normal distributions. Computational results on an IBM 3090 computer are reported for a set of test problems.
- Research Organization:
- Indiana Univ., Bloomington, IN (United States)
- OSTI ID:
- 5313055
- Resource Relation:
- Other Information: Thesis (Ph.D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
A parallel genetic algorithm for the set partitioning problem
SU-F-BRD-13: Quantum Annealing Applied to IMRT Beamlet Intensity Optimization