Verification and extension of the theory of global-local hybrids

This work is an extension of the framework for optimizing global-local hybrids. The existing theory idealizes the search problem as a search by a global searcher for acceptable targets or for basins of attractions which lead to acceptable target by invoking a local searcher. The two key parameters of this theory are—the probabilities of successfully hitting targets and basins and time-to-criterion values for different basins. First the existing theory is tested with variation in time-to-criterion values for the local searcher across several basins and is then extended to handle variations within individual basins. As a first step towards applying this theory to genetic algorithms (as the global searcher), selection dominated performance has also been studied in the context of this theory. The results are promising and make a strong case for further work in this direction.