Case-by-Case Problem Solving
Pei Wang · 2009
Case-by-case Problem Solving solves each occurrence, or case, of a problem using available knowledge and resources on the case.It is different from the traditional Algorithmic Problem Solving, which applies the same algorithm to all occurrences of all problem instances.Case-by-case Problem Solving is suitable for situations where the system has no applicable algorithm for a problem.This approach gives the system flexibility, originality, and scalability, at the cost of predictability.This paper introduces the basic notion of Case-by-case Problem Solving, as well as its most recent implementation in NARS, an AGI project. Algorithmic Problem Solving"Problem Solving" is the process to find a solution for a given problem by executing some operations.For a certain system at a certain moment, the set of executable operations usually remains constant.Therefore, the task for the system is to find a way to select proper operations and to execute them in proper order for the given problem.In computer science and AI, the dominant approach in problem solving can be called "Algorithmic Problem Solving" (APS in the following).According to this approach, first a problem is specified as a function that maps any input (problem instance) of a certain type to the corresponding output.Then, an algorithm is designed, which accomplishes this function step by step, where each step is a well-defined operation.Finally, the algorithm is implemented in a programming language to become a computer program, which will be able to let a computer routinely transform valid input data into output data.A well-known description of this approach can be found in (Marr, 1982).Accurately speaking, in this approach "problem solving" happens in two different levels:1.When the problem refers to a problem type, or inputoutput mapping, the solution is the corresponding algorithm (conceptually speaking) or program (practically speaking) that accomplishes the mapping.For example, when the problem is "to sort sequences of comparable items", one solution is "quicksort".