Genetic
algorithms model biological evolution to solve a variety
computational problems. We investigated enhancements to
the traditional genetic algorithm suited for difficult global
optimization problems, including the use of parallel computers
to speedup the execution and improve the quality of solution.
Based on our studies, we designed a configurable, parallel
genetic algorithm package that offers the choice of three
parallel genetic algorithm models: global, island, and neighborhood.
We configured and applied a parallel genetic algorithm to
a difficult global optimization problem from the field of
material science: structure determination from Low Energy
Electron Diffraction experiments.
