IRI-9503366 Whitley, L. Darrell Colorado State University $90,000 - 12 mos. Comparisons and Applications of Local and Global Search This is first year finding for a three year continuing grant. Scientists and engineers seeking practical ways to solve difficult combinatorial optimization problems are increasingly turning to local search and genetic algorithms. In addition, the past five years has seen a dramatic increase in researchers proposing new variants on these techniques, especially genetic algorithms. Recent work on genetic algorithms has strongly suggested the best algorithms interweave local search and genetic search techniques. This research has four main objectives: 1) produce a suite of local search methods and stochastic hill-climbing algorithms to test the difficulty of optimization problems, 2) produce a new test suite of complex problems which are resistant to local search and stochastic hill-climbing, 3) produce new hybridized genetic algorithms that can be sued to produce state-of-the-art results for applications in computer vision and geophysics, 4) conduct comparative studies of the major genetic algorithm search paradigms, including hybrids which utilize local search. In pursuit of this fourth objective, it will be necessary to compare different genetic algorithms, local search algorithms, and hybrids on both the controlled test suite of problems as well as large problems of practical importance.
