SWO and local search

Although the ability to make large, coherent moves is a strength of the approach, it is also a weakness. SWO is poor at making small ``tuning'' moves in the solution space, but the coupled-search view of SWO suggests an obvious remedy. SWO could be combined with local search in the solution space, to look for improvements in the vicinity of good solutions. Similarly, making small changes to a prioritization would generally result in smaller moves in the solution space than result from going through the full analysis and re-prioritization cycle.

Yet another alternative is genetic algorithm techniques for ``crossover'' and other types of mutation to a pool of nodes, as is done in OPTIFLEX. Many hybrid approaches are possible, and we believe that the coupled-search view of SWO helps to identify some interesting strategies for combining moves of various sizes and kinds, in both search spaces, adapting dynamically to relative solution qualities.