区间二型模糊子空间0阶TSK系统

人们倾向于使用少量的有代表性的特征来描述一条规则,而忽略极为次要的冗余的信息。经典的区间二型TSK(Takagi-Sugeno-Kang)模糊系统,在规则前件和后件部分会使用完整的数据特征空间,对于高维数据而言,易导致系统的复杂度增加和可解释性的损失。针对于此,提出了区间二型模糊子空间0阶TSK系统。在规则前件部分,使用模糊子空间聚类和网格划分相结合的方法生成稀疏的规整的规则中心,在规则后件部分,使用简化的0阶形式,从而得到规则语义更为简洁的区间二型模糊系统。在模拟和真实数据上的实验结果表明该方法分类效果良好,可解释性更好。     

可解释的深度TSK模糊系统综述

深度神经网络在多个领域取得了突破性的成功,然而这些深度模型大多高度不透明。而在很多高风险领域,如医疗、金融和交通等,对模型的安全性、无偏性和透明度有着非常高的要求。因此,在实际中如何创建可解释的人工智能（Explainable artificial intelligence, XAI）已经成为了当前的研究热点。作为探索XAI的一个有力途径,模糊人工智能因其语义可解释性受到了越来越多的关注。其中将高可解释的Takagi-Sugeno-Kang（TSK）模糊系统和深度模型相结合,不仅可以避免单个TSK模糊系统遭受规则爆炸的影响,也可以在保持可解释性的前提下取得令人满意的测试泛化性能。本文以基于栈式泛化原理的可解释的深度TSK模糊系统为研究对象,分析其代表模型,总结其实际应用场景,最后剖析其所面临的挑战与机遇。     

模糊系统和神经网络的特征与比较

概述了模糊、神经网络 和人工智能技术之间的关系，尤其探讨了模糊系统和神经网络的特性；指出了模糊系统和神经网络的结合方式，分析了它们的特征。     

基于模糊神经网络的智能火灾报警系统

针对目前建筑物发生火灾事故后果严重、损失大的问题,本文主要设计了基于模糊神经网络的智能火灾报警系统.采用模糊控制理论可以提高报警系统的灵敏度,减少误报率;结合神经网络的自学习功能可以提高整个系统的智能化程度.整个火灾报警系统采用分布智能型结构.     

基于模糊划分和支持向量机的TSK模糊系统

为了提高模糊系统处理高维问题的推广能力与鲁棒性能,提出将模糊聚类和支持向量机算法结合起来构造TSK模糊系统的算法.首先运用模糊聚类算法对输入空间进行划分,确定模糊规则前件的隶属函数.然后用支持向量机算法确定模糊规则的后件参数.该支持向量机的核函数是由模糊规则前件的隶属函数构造的,并且是Mercer核.在3个数据集的实验结果表明,与TSK模糊系统的传统算法和支持向量机相比较,本文算法具有更好的推广能力和鲁棒性.     

模糊系统与前向神经网络的结合

1 引言模糊系统方法和神经网络技术是近年来计算智能领域研究热点,被广泛地应用于复杂系统、非确定性等难于建立比较准确的数学模型的问题,并在自动控制、计算机图像处理、语音识别、手写体识别等领域有重要应用。模糊系统与神经网络的结合也越来越受到人们的重视。模糊系统和神经网络的结合可以分为模糊系统与前向网络的结合和与反馈网络的结合两类。模糊系统与反馈网络的结合主要有模糊联想记忆、模糊     

1996年《中国无线电电子学文摘》收录《数据采集与处理》文摘情况

１９９６年《中国无线电电子学文摘》收录《数据采集与处理》文摘情况《中国无线电电子学文摘》是由中国科学院文献情报中心和中国科学院电子学研究所情报室联合主办，国家教委批准的国家一级检索刊物，专门收集报道我国（包括港、澳、台）广大科技工作者在国内和港台地区...     

模糊CMAC神经网络用于MIMO非线性系统的反馈线性化

针对一类多输入多输出（ＭＩＭＯ）连续时间非线性系统,应用模糊ＣＭＡＣ神经网络,给出一种状态反馈控制器,用于使状态反馈可线笥化的未知的非线性动态系统儿得要求的患 很弱的假设条件下,应用李雅普诺夫稳定性理论严格地证明了闭环系统内的所有信号为一致最终有界（ＵＵＢ）。     

Stable Adaptive Fuzzy Control with TSK Fuzzy Friction Estimation for Linear Drive Systems

This paper considers the control of a linear drive system with friction and disturbance compensation. A stable adaptive controller integrated with fuzzy model-based friction estimation and switching-based disturbance compensation is proposed via Lyapunov stability theory. A TSK fuzzy model with local linear friction models is suggested for real-time estimation of its consequent local parameters. The parameters update law is derived based on linear parameterization. In order to compensate for the effects resulting from estimation error and disturbance, a robust switching law is incorporated in the overall stable adaptive control system. Extensive computer simulation results show that the proposed stable adaptive fuzzy control system has very good performances, and is potential for precision positioning and trajectory tracking control of linear drive systems.     

基于模块化模糊神经网络的非线性系统故障诊断

提出了一种基于模块化模糊神经网络的非线性系统故障诊断新方法。该方法先使用模糊c-均值聚类法对测量空间进行模块分割,再利用模糊IF－THEN规则对分割后的子空间分别采用局部BP模型进行逼近,最后,通过离线学习获得不同子空间故障输出与测量输入的非线性动力特性。试验表明该网络具有良好的泛化性能,可显著提高非线性系统故障检测的快速性、鲁棒性及准确率。     

多变量系统的模糊神经网络控制模型及其应用

本文综合模糊控制系统与人工神经网络的优点，提出了一种多变量系统的模糊神经网络控制模型并给出了其建模方法，该方法适合于多变量系统的建模及其模糊控制器的设计。笔者以此方法建立了渣贫化电炉生产过程的模糊神经网络控制模型并开发出相应的决策支持系统，该系统自１９９２年６月投入生产现场使用以来，一直稳定可靠地正常运行，取得了令人满意的效果和显著的经济和社会效益。     

具有时滞的不确定性系统神经网络模糊自学习控制

本文对具有时滞的不确定性控制对象提出了一种神经网络时滞补偿模糊自学习控制方法．模糊控制器采用误差、误差变化及误差加速度的加权和的解析描述形式，利用人工神经网络直接对过程建模，实现对时滞补偿预报以及对模糊加权因子的自学习优化调整．将上述方法用于焊接熔池动态过程控制试验，结果表明本文提出的自学习神经网络时滞补偿模糊控制方案有效．     

一种自组织模糊神经网络控制器

采用一种具有结构和参数学习能力的自组织模糊神经网络控制器设计方法。这种控制器无需事先确定模糊控制规则，能在控制过程中通过神经网络的结构及参数学习在线调整模糊神经网络的结构、产生模糊控制规则、调整规则的参数。仿真表明该控制器能用于一定纯滞后时变对象的控制，具有良好的控制性能。     

改进型B样条模糊神经网络

提出了一种改进型B样条模糊神经网络,用以增加B样条函数在作为模糊神经网络控制器实际应用中的抗干扰能力,通过指出常规B 样条模糊隶属函数构造中存在的不足,提出了针对性的改进措施,从而既满足了B样条本身的性质,又从理论设计上避免了使系统不稳定的因素,在实际电机速度控制系统的对比实验中证明了所提改进方法的有效性和实用性,文中给出了改进型B样条网络的设计过程及实验结果。     

鲁棒性的模糊聚类神经网络

针对模糊聚类神经网络FCNN(fuzzy clustering neural network)对例外点(outliers)敏感的缺陷,通过引入Vapnik's ε-不敏感损失函数,重新构造网络的目标函数,并根据拉格朗日优化理论推导出新的鲁棒模糊聚类神经网络及其算法(robust fuzzy clustering neural networks,简称RFCNN).RFCNN有效地克服了FCNN对例外点敏感之缺点并且能得到合理的聚类中心.仿真实验结果表明,RFCNN较之于FCNN有更好的鲁棒性.     

Fuzzy Sets and Neural Networks

Abstract

It is possible that a better model for the behavior of a nerve cell may be provided by what might be called a fuzzy neuron, which is a generalization of the McCulloch-Pitts model. The concept of a fuzzy neuron employs some of the concepts and techniques of the theory of fuzzy sets which was introduced by Zadeh [2, 3] and applied to the theory of automaton by Wee and Fu [6], Tanaka et al. [7], Santo [8] and others. In effect, the introduction of fuzziness into the model of a neuron makes it better adapted to the study of the behavior of systems which are imprecisely defined by virtue of their high degree of complexity. Many of the biological systems, economic systems, urban systems and more generally, large-scale systems fall into this category.

In the nearly three decades since its publication, the pioneering work of McCulloch and Pitts [1], has had a profound influence on the development of the theory of neural nets, in addition to stimulating much of the early work in automata theory and regular events.

Although the McCulloch-Pitts model of a neuron has contributed a great deal to the understanding of the behavior of neural-like systems, it fails to reflect the fact that the behavior of even the simplest type of nerve cell exhibits not only randomness but, more importantly, a type of imprecision which is associated with the lack of sharp transition from the occurrence of an event to its non-occurrence.

In this paper, some basic properties of fuzzy neural networks as well as their applications to the synthesis of fuzzy automata are investigated. It is shown that any n-state minimal fuzzy automaton can be realized by a network of m fuzzy neurons, where ┌log₂ n┐ ? m ? 2n. Examples are given to illustrate the procedure. As an example of application, a realization of λ-fuzzy language recognizer using a fuzzy neural network is presented. The techniques described in this paper may be of use in the study of neural networks as well as in formal languages, pattern recognition, and learning.     

Artificial Neural Networks Manipulation Server: Research on the Integration of Databases and Artificial Neural Networks

This paper proposes a new whole and distributed integration approach between Artificial Neural Networks (ANNs) and Databases (DBs) taking into account the different stages of the former’s lifecycle (training, test and running). The integration architecture which has been developed consists of an ANN Manipulation Server (AMS) based on a client-server approach, which improves the ANNs’ manipulation and experimentation capabilities considerably, and also those of their training and test sets, together with their modular reuse among possibly remote applications. Moreover, the chances of integrating ANNs and DBs are analysed, proposing a new level of integration which improves the integration features considerably. This level has not been contemplated yet at full reach in any of the commercial or experimental tools analysed up to the present date. Finally, the application of the integration architecture which has been developed to the specific domain of Environmental Impact Assessments (EIAs) is studied. Thus, the versatility and efficacy of that architecture for developing ANNs is tested. The enormous complexity of the functioning of the patterns which rule the environment’s behaviour, and the great number of variables involved, make it the ideal domain for experimenting on the application of ANNs together with DBs.     

一个基于神经网络的信息系统安全性综合评估模型

本文提出了一个将神经网络技术与模糊综合评价法结合的评估模型。首先根据信息系统资产的组成以及安全性因素建立层次性的安全性指标体系,使用层次分析法确定指标权重;然后借助安全工具的测试结果,使用模糊综合评估法构造前向神经网络;最后,使用神经网络的反向传播算法调整指标权重。在此模型的基础上,设计并实现了一个信息息系统安全性评估系统。     

人工神经网络的符号解释

提出了人工神经网络符号解释的基本过程，并详细阐述了该过程中的三个重要步骤：网络的构建、规则的提取以及规则的评估。     

人工神经网络在滥用检测上的应用

目前大多数滥用检测(misusedetection)的方法,是使用基于规则的专家系统来鉴别已知攻击的种种迹象的,然而,一旦攻击模式改变,这些技术就很难识别出来,而人工神经网络则有着非常大的潜力,可以基于有限、不完整和非线形的数据资源对网络行为进行鉴别和分类。