Adaptive Problem Solving
Abstract
Artificial Intelligence has created a large repertory of tools (e.g., algorithms, representational schemes, and heuristics) for solving problems. However, no one tool choice dominates all others. Similarly, given a problem, there is no automatic way to pick an appropriate mix of tools to efficiently solve that problem. This is the problem this research is tackling. We are developing an architecture, that given a problem, looks at its repertory of tools and current computing environment and predicts the impact of different alternatives upon solution time. Based on these predictions, it selects the most appropriate alternatives. It monitors its progress against its predictions to decide whether it should reassess the suitability of its choices and possibly alter them.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 23, 2016
- Source ID
- FA23861514069
Entities
People
- Michael W. Barley
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Auckland