The Consensus of Uncertainties in Distributed Expert Systems

Minjie Zhang and Chengqi Zhang

This paper deals with the consensus of uncertainties in distributed expert systems (DESs). It claims that the different uncertainties of a solution from different expert systems constitute not only a conflict case, but also non-conflict cases. We will refer both the conflict case and non-confiict cases as synthesis cases. The first objective of this paper is to classify the synthesis cases, identify the types of DESs, and recognize the relationships among the different synthesis cases and the different types of DESs. On the basis of this, a computational synthesis strategy is proposed to obtain a consensus of uncertainties in a conflict case. Within this strategy, the conflict case is further classified into two sub-cases, and two corresponding sub-strategies are proposed, in which both uncertainties and authorities are taken into consideration. A synthesis strategy based on neural networks is also proposed in a conflict case. In this strategy, as long as enough patterns have been obtained from human experts, neural networks can be trained to match all patterns. This strategy can simulate human experts reasonably well. Tests have also shown that a fixed neural network architecture can be used to solve conflict problems, with a variable number of inputs and outputs. That means only a small number of neural networks are required to solve all conflicts, thus the neural network can be used in real DESs. Finally, a computational synthesis strategy is compared to a neural network synthesis strategy. Both strategies have advantage and disadvantage and are adapt to different situations.


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.