Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
9651748 | International Journal of Approximate Reasoning | 2005 | 19 Pages |
Abstract
This paper addresses the robustness characteristics of the fuzzy inference mechanism in terms of maximum deviation of the fuzzy and crisp output as a result of the deviation of the input membership grades. A formulation that introduces several parameters into the fuzzy reasoning process provides a suitable means to adjust the robustness of the inference engine. The effect of each of these parameters is investigated and specific guidelines for assigning their range are developed to achieve maximum robustness. The maximum possible robustness is achieved by reducing the sensitivity of the inference mechanism to input variation to a satisfactory level. This feature will improve the generalization capability of fuzzy-logic models as illustrated with a well-known example from the literature.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
William W. Melek, Andrew A. Goldenberg,