The sensitivity properties of hierarchical logic-based models

摘要

Much of whe work on logic modeling focuses on constructing logic-based models that represent real-world situations, investigating the properties of different types of models and logics, and developing solution procedures for the models so as to address real-world problems. An important subset of the latter task is to perform sensitivity analyses that determine how the output of such a model will change in response to a change in its inputs. In two previous papers we developed methods for calculating sensitivity measures for Boolean models using the concept of a Boolean derivative, initially developed for circuit design. In this paper we review these methods and present a method for calculating sensitivity measures, in the form of first derivatives, for models based on fuzzy logic. Since these models are piecewise linear, it is necessary to consider derivatives from both the right and the left.