homeresources & publications › modeling when connections are the problem


Modeling when connections are the problem


Most of the AI diagnostic reasoning approaches for digital systems presume that digital components can be modeled as pure functions of their inputs, all signals can be represented by ``1''s and ``0''s, wires between components cannot fail, and do not model replacement of components or wires. None of these assumptions are valid for the challenges diagnosticians encounter in real systems. This paper presents a digital expert (DEX) that can reason over digital systems as an electrical engineer would: at a qualitative, causal level more accurate than the simple ``0''/``1'' level, but without incurring the costs of full-scale numerical algorithms. A fundamental contribution of this paper is a more powerful approach to modeling connections which does not require special-case post-processing and is computationally tractable.


de Kleer, J. Modeling when connections are the problem. Twentieth International Joint Conference on Artificial Intelligence (IJCAI-2007); 2007 January 6-12; Hyderabad; India; 310-317.