A heuristic for decomposing a problem into a sequence of subproblems.

Abstract

This thesis presents a method for decomposing a specification of a problem into a sequence of subproblem specifications. The method uses the specification to build a tree-like structure called a semantic net. The net is then used to construct a sequence of subspecifications. Each subspecification of the sequence represents a subproblem. Composition of the solutions of the subproblems results in a solution to the given problem specification. In this work, we present an intuitive approach to what Artificial Intelligence and program synthesis is, define the sequence problem associated with program synthesis, and present the method for deriving a sequence of subspecifications. When this has been done, the method is then applied to a specific problem domain called the Blocks World. We then consider the method in a non-Blocks World domain and follow with a summary.