A comparative study on methods for reducing myopia of hill-climbing search in multirelational learning