Approximation Reformulations

Daniel S. Weld

Although computers are widely used to simulate complex physical systems, crafting the underlying models that enable computer analysis remains difficult. When a model is created for one task, it is often impossible to reuse the model for another purpose because each task requires a different set of simplifying assumptions. By representing modeling assumptions explicitly as approximation reformulations, we have developed qualitative techniques for switching between models. We assume that automated reasoning proceeds in three phases: 1) model selection, 2) quantitative analysis using the model, and 3) validation that the assumptions underlying the model were appropriate for the task at hand. If validation discovers a serious discrepancy between predicted and observed behavior, a new model must be chosen. We present a domain independent method for performing this model shift when the models are related by an approximation reformulation and describe a Common Lisp implementation of the theory.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.