Timothy L. Hinrichs, Michael R. Genesereth
Computer scientists routinely design algorithms to efficiently solve problems found in nature. The resulting algorithms encapsulate the original problem as well as extra information about how to solve that problem; thus, nature's original description that says what the problem is has been augmented with information about how to solve it. In this paper, we investigate the automation of this process by concentrating on declarative languages, arguing that certain classes of declarative languages encode more information about how to solve a problem than others, and demonstrating techniques for automatically translating between two languages separated in the what-to-how spectrum.
Subjects: 3. Automated Reasoning; 15. Problem Solving
Submitted: Jun 15, 2008