This Is AuburnElectronic Theses and Dissertations

A Framework for Universal Problem Resolution with Continuous Improvement

Date

2014-05-01

Author

Leithiser, Robert

Type of Degree

dissertation

Department

Computer Science

Abstract

This dissertation outlines a universal problem resolution framework implementing reasoning and improvement capabilities exhibited by human beings. The work reviews literature concerning major differentiations of human learning compared to contemporary artificial intelligence (A/I) systems. Deductions from the body of knowledge in mathematics, computer science, and human learning quantify the human capability for continuous and recursive self-reflection as a central differentiator. Analysis of capabilities of computational systems establishes that structural differences between the human brain and computational devices do not preclude computational implementation of human self-reflective capability. The work establishes self-reflective capability as an enabler for a co-recursive and recursive paradigm operable with unlimited depth and breadth for the problem of optimizing problem solving itself. Discovering the general algorithm for the Tower of Hanoi is the base case used to show how simulation outputs can transform to higher order pattern recognition problems. The Tower of Hanoi provides a model that is common for any problem in that it includes an initial state, a desired outcome state, and an allowed transition state (where states allow multiple combinations of sub-states). This model is representable in the system and thus enabled for solution discovery. Detailed examples wherein the system pursues problem search spaces in a manner enabling autonomous self-improvement as a natural result of encountering new types of problems validate the thesis. Cooperating agents analyze solution path determinations for problems including those concerning their own optimization. This spawns state transition rules generalizable to higher layers of abstraction resulting in new knowledge enabling self-optimization. Major outcomes include: 1) Proof by example that a continuous improvement universal problem resolution framework is constructible using currently available software and hardware, 2) Production of a prototype that meets the thesis for a continuous improvement system – non-domain specific, extensible without reprogramming of the core system, lacking need of subject matter expert intervention; and executing within polynomial time complexity, 3) Rationale that the framework coupled with technology advancements generate optimal solutions using fewer resources than possible without such a framework.