一种基于模糊神经网络加权的多维稀疏模糊推理方法

Multidimensional Fuzzy Reasoning Method Based on Weight of Fuzzy Neural Network

在稀疏规则库条件下,经典的插值理论针对一维稀疏规则库提出了各种不同的插值方法,取得了很多很好的经验;但对多维稀疏规则条件的近似推理研究很少,不仅存在着难以保证推理结果的凸性和正规性等问题,而且没有考虑到多维变量之间的联系即对结论的影响权值,造成推理结果的误差性更大。多变量规则的模糊插值推理是插值推理研究的重要方面,为了在多变量稀疏规则条件下得到好的插值推理效果,本文提出了一种基于模糊神经网络加权的多维模糊推理方法,为智能系统中的模糊推理提供了一个十分有用的工具。

Interpolative reasoning is type of important reasoning approaches under sparse rules. Interpolative reasoning in one dimension has been researched widely, but the research in multi-dimension is lacking and a few existing approaches have some faults. These methods not only cannot guarantee the convexity of result, but also cannot consider the relation between many variables, weight of influenced conclusion. It leads to more error of inferential result. Interpolative reasoning in multi-dimension is an important research aspect of interpolative reasoning, in order to get better conclusion under multidimensional sparse rules condition, we propose a fuzzy multidimensional reasoning method based on weight of fuzzy neural network, which moreover can keep the convexity of the reasoning consequence.