My research is focused on parameter optimization for search algorithms, such parameters control how algorithms traverse the search space, and greatly affect both the resulting solution quality and running time. While experienced users can take advantage of such parameters, most people will just use the default values, or use a trial-and-error approach to tune the algorithm, which is a very tedious task that is not likely to produce good results, and will have to be repeated for different problem domains, instances, or even the same instance if the problem is dynamic.

Informally speaking, the problem can be defined as: finding parameter values for which an algorithm reaches its best possible performance. “Best performance” depends on the application, it could be, for example, the objective function value, the time needed to reach a target solution, or the number of problem instances, or domains, an algorithm can solve.