Search: from Algorithms to Systems

  • Youssef Hamadi

MSR-TR-2012-89 |

To conclude the presentation of our research, we present the concept of Autonomous Search in Chapter 6. We de fine autonomous solvers as solvers that contain control in their search process, and study such autonomous systems w.r.t. their speci fic control methods. A control process includes a strategy that manages the modi fication of some of the solver’s components and behavioral features after the application of some solving functions. The overall strategy to combine and use components and parameters can be based on learning that uses information from the current solving process and/or from previously solved instances. This chapter proposes a taxonomy of search processes w.r.t. their computation characteristics, and provides a rule-based characterization of autonomous solvers. This allows a formalizing of solvers adaptations and modifications with computation rules that describe the modi fication of the solver’s components transformation.