Although the computation power grows rapidly, finding solutions for search problems remain an unrealistic option for many hard domains. This is why the field of non-optimal search guided by heuristic functions is crucial for solving practical problems. In my research I use Evolutionary Algorithms (EAs) techniques, inspired by Darwin’s theory, to develop hyper-heuristics for guiding search.

The Method

I have designed a novel approach for evolving compound heuristic functions based on simpler building blocks [publications]. These building blocks can be either 1) domain independent heuristic, 2) domain specific fast and easy-to-implement heuristics, or 3) memory based / complex heuristics designed specifically for the domains in question.

Test Cases

So far I’ve tested my method on the Rush Hour puzzle and the game of FreeCell. My results have shown a significant reduction in search time and computation resources compared to other state-of-the-art solvers for these domains as well as human players. Thus, EA outperform both human solvers of the domain in question and also human designers of heuristics.

My solvers won in two international competitions and I have received the prestigious Faran scholarship for outstanding Ph.D. students.

I have developed a framework for evolving hyper-heuristics called – HH-Evolver.

Here are related post:

[catlist name=”research”]