多重加权多值指数双向联想记忆网络及其表决性能

Wang和陈等利用各自提出的二值指数双向联想记忆模型（eBAM)及其改进型eBAM(IeBAM),分别构造了由多个eBAM和IeBAM组成的多重eBAM(Multi-eBAM)和多重IeBAM(Multi-IeBAM)的信念组合模型,使之可模拟多个专家的表决。该文在此基础上,借助陈提出的多值eBAM(MVeBAM),提出了多重多值eBAM (Multi-MVeBAM),对Multi-eBAM和Multi-IeBAM进行了两方面的推广;一是将二值表示推广到多值表示,以此可以处理现实中的多值数据;二是将原有模型中具有同等权威度的各专家推广到各具不同的权威度的专家,以此模拟更实际的表决情形。文中借助能量函数证明了所提模型的渐近稳定性,以保证其实际可用。计算机模拟证实了模型的可行性。     

不完备信息下基于切换推理证据网络的空中目标识别方法

现有证据推理方法模型结构固定、信息处理方式和推理机制单一，难以适用于集结了不确定、错误甚至缺失等多种不完备信息环境下的目标识别。针对该问题，提出了一种切换推理证据网络（SR-EN）方法。首先，考虑证据节点删除等情况构建多模板网络模型；然后，分析各证据变量与目标类型的条件关联性以建立针对不完备信息的推理规则库；最后，提出基于三种证据输入及修正方式的智能化时空融合推理方法。与传统的证据网络（EN）以及EN与优劣解距离法（TOPSIS）等两种信息修正方法的结合方法相比，SR-EN能够在确保推理时效性的同时实现在多类随机性不完备信息下对空中目标的连续准确识别。实验结果表明，通过对各类不完备信息的有效识别，SR-EN能够实现连续推理过程中证据处理方式、网络结构和节点间融合规则的自适应切换。     

基于规则推理网络的分类模型

为了缓解神经网络的“黑盒子”机制引起的算法可解释性低的问题,基于使用证据推理算法的置信规则库推理方法(以下简称RIMER)提出了一个规则推理网络模型.该模型通过RIMER中的置信规则和推理机制提高网络的可解释性.首先证明了基于证据推理的推理函数是可偏导的,保证了算法的可行性;然后,给出了规则推理网络的网络框架和学习算法,利用RIMER中的推理过程作为规则推理网络的前馈过程,以保证网络的可解释性;使用梯度下降法调整规则库中的参数以建立更合理的置信规则库,为了降低学习复杂度,提出了“伪梯度”的概念;最后,通过分类对比实验,分析了所提算法在精确度和可解释性上的优势.实验结果表明,当训练数据集规模较小时,规则推理网络的表现良好,当训练数据规模扩大时,规则推理网络也能达到令人满意的结果.     

改进证据推理的联合防空作战效能评估模型

根据证据理论在处理不确定性推理问题时的优势,运用证据理论对联合防空作战效能进行评估。首先建立基于证据推理的基本模型,然后引入模糊数学方法来处理具有模糊概念或推理关系的复杂问题,并且还考虑了实际问题中可能出现的加权证据或者相关证据的情况,提出了利用改进的证据模型进行联合防空作战效能评估的方法。通过实例分析,证明了改进证据推理模型在联合防空作战效能评估中的实用性与有效性。     

基于证据推理网络的实时网络入侵取证方法

在分析现有网络入侵取证系统所存在问题的基础上,提出了一种基于证据推理网络的实时网络入侵取证方法NetForensic,将弱点关联性的概念引入网络入侵取证领域,根据网络系统的弱点知识和环境信息构建了证据推理网络,利用证据推理网络所提供的多阶段攻击推理能力,NetForensic实现了高效实时攻击流程重构.实验数据表明,NetForensic给出的证据链完整可信,且具备实时推理的能力,为快速有效的调查取证提供了有力保证.     

改进的指数双向联想记忆模型及性能估计

提出了一个新的改进型指数双向联想记忆模型(improved eBAM,简称IeBAM).通过定义有界且随状态改变而下降的能量函数,证明了IeBAM在状态的同、异步更新方式下的稳定性,一方面排除了Wang的修正指数BAM(modified eBAM,简称MeBAM)和Jeng的eBAM(exponential BAM)的稳定性证明中所作的不合理假设;另一方面,放宽了对BAM(bidirectional associative memory)的连续性假设的要求,并避免了补码问题.理论分析和计算机模拟结果表明,     

一种基于条件证据网络的多源异类知识融合识别方法

针对融合识别领域中不同框架下多源异类传感器的不确定证据信息无法有效融合的问题, 提出一种基于条件证据网络的多源异类知识融合识别方法. 该方法将战场协同作战中不同框架下多源异类传感器的领域知识统一在证据网络的结构下, 形成多源异类知识融合识别模型, 对多源异类传感器的不确定性证据信息进行基于条件证据网络的融合推理, 得到识别结果. 仿真实例验证了所提出方法的优越性.

     

基于贝叶斯网络的多传感器目标识别算法研究

基于贝叶斯网络能够组合多种证据进行不确定性表达和推理的特点,提出以贝叶斯网络为基本结构的目标融合识别模型.通过详细分析空中目标识别的推理规则,建立了空中目标识别的贝叶斯网络拓扑结构.首先对各传感器的数据分别进行融合,然后应用贝叶斯网络推理算法对多种传感器融合结果进行融合计算,最后根据假定变量各状态的概率取值来判断目标平台类型.仿真结果证明了该方法直观、形象,计算速度快,降低了实用的复杂度,提高了目标识别的可靠性.     

基于TBM模型的多Agent决策融合方法

多Agent系统存在的动态特性使证据推理中的可传递置信模型（TBM）能够有效地处理动态环境的证据推理。在分析和研究可传递置信模型算法的基础上,提出一种基于证据推理TBM模型的多Agent决策融合方法,构建多Agent决策融合系统的框架模型,分析该系统的信息更新、合成算法及决策制定算法。利用SimuroSot作为仿真平台,将该方法应用于判断对手的队形和策略,得到了较满意的结果。     

基于改进模糊熵和证据推理的多属性决策方法

针对证据推理方法框架下属性权重难以获取的问题,提出一种基于改进模糊熵和证据推理的多属性决策方法。首先,定义证据推理信度决策矩阵框架下的三角函数模糊熵公式,并证明了其满足熵的四个公理化定义。其次,所提方法能够同时处理属性权重完全未知和属性权重信息部分已知两种情况：当属性权重完全未知时,基于信度框架下的改进模糊熵和熵权法的基本思想计算属性权重;当属性权重信息部分已知时,定义加权模糊熵,建立期望模糊熵最小的线性规划模型求解最优属性权重。最后,利用证据推理算法融合方案属性值,结合期望效用理论得到方案排序结果。通过实例计算,并与传统模糊熵计算方法进行比较分析,验证了所提方法能够更加充分地反映原始决策信息,更具客观性和一般性。     

The majority theorem of centralized multiple BAMs networks

A method for modeling the learning of belief combination in evidential reasoning using a neural network is presented. A centralized network composed of multiple bidirectional associative memories (BAMs) sharing a single output array of neurons is proposed to process the uncertainty management of many pieces of evidence simultaneously. The convergence properties of the multi-BAM network are proved. The combination process of evidence is considered as a resonant process through the multi-BAM networks. Most important of all, a majority rule of decision making in presentation of multiple evidence is also found by the study of signal-noise-ratio (SNR) of the multi-BAM network. Some simulation examples are given. The result is coherent with the intuition of reasoning.     

The decision-making properties of discrete multiple exponentialbidirectional associative memories

A method for modeling the learning of belief combination in evidential reasoning using a neural network is presented. A centralized network composed of multiple exponential bidirectional associative memories (eBAM's) sharing a single output array of neurons is proposed to process the uncertainty management of many pieces of evidence simultaneously. The stability of the proposed multiple eBAM network is proved. The sufficient condition to recall a stored pattern pair is discussed. Most important of all, a majority rule of decision making in presentation of multiple evidence is also found by the study of signal-noise-ratio of multiple eBAM network. A guaranteed stable state condition, i.e., the condition for the fastest recall of a pattern pair, is also studied. The result is coherent with the intuition of reasoning.     

Multiple behavior information fusion based quantitative threat evaluation

How to evaluate network security threat quantitatively is one of key issues in the field of network security, which is vital for administrators to make decision on the security of computer networks. A novel model of security threat evaluation with a series of quantitative indices is proposed on the analysis of prevalent network intrusions. This model is based on multiple behavior information fusion and two indices of privilege validity and service availability that are proposed to evaluate the impact of prevalent network intrusions on system security, so as to provide security evolution over time, i.e., monitor security changes with respect to modification of security factors. The Markov model and the algorithm of D-S evidence reasoning are proposed to measure these two indices, respectively. Compared with other methods, this method mitigates the impact of unsuccessful intrusions on threat evaluation. It evaluates the impact of important intrusions on system security comprehensively and helps administrators to insight into intrusion steps, determine security state and identify dangerous intrusion traces. Testing in a real network environment shows that this method is reasonable and feasible in alleviating the tremendous task of data analysis and facilitating the understanding of the security evolution of the system for its administrators.     

A novel fuzzy evidential reasoning paradigm for data fusion with applications in image processing

This paper presents a novel data fusion paradigm based on fuzzy evidential reasoning. A new fuzzy evidence structure model is first introduced to formulate probabilistic evidence and fuzzy evidence in a unified framework. A generalized Dempster’s rule is then utilized to combine fuzzy evidence structures associated with multiple information sources. Finally, an effective decision rule is developed to take into account uncertainty, quantified by Shannon entropy and fuzzy entropy, of probabilistic evidence and fuzzy evidence, to deal with conflict and to achieve robust decisions. To demonstrate the effectiveness of the proposed paradigm, we apply it to classifying synthetic images and segmenting multi-modality human brain MR images. It is concluded that the proposed paradigm outperforms both the traditional Dempster–Shafer evidence theory based approach and the fuzzy reasoning based approach     

实体异构性下证据链融合推理的多属性群决策

针对多属性群决策中可解释性证据融合推理的实体异构性问题,给出了一个实体异构性下证据链融合推理的多属性群决策方法.基于证据推理理论,引入证据链关联的概念,从多数据表提供的数据矩阵中获取可区分的近邻证据集,推导了各数据表的相似度矩阵,并构建半正定矩阵的二次优化模型,共享群决策专家的经验知识.使用Dempster正交规则,论证了异构实体之间可解释性推理中可信度融合的合理性,并使用证据融合规则集成各个数据表的近邻证据中获得的可信度,验证了调和多源异构数据中不一致信息的有效性.通过具有实体异构性的心脏病多决策数据诊断实例说明了方法的可行性与合理性.     

D-S证据理论在决策支持系统中的应用

D-S证据理论提供了一种解决多数据源不确定信息推理和融合的有效方法。证据理论能够对各自独立的证据加以综合给出一致性结果,并能处理具有模糊和不确定信息的合成问题,最终达到信息互补。与其他推理方法相比更符合人类思维决策过程。为此,提出一种基于D-S证据理论的灾害决策支持方法,并根据试验结果验证了该方法的有效性和可行性。     

On the evidential reasoning algorithm for multiple attribute decision analysis under uncertainty

In multiple attribute decision analysis (MADA), one often needs to deal with both numerical data and qualitative information with uncertainty. It is essential to properly represent and use uncertain information to conduct rational decision analysis. Based on a multilevel evaluation framework, an evidential reasoning (ER) approach has been developed for supporting such decision analysis, the kernel of which is an ER algorithm developed on the basis of the framework and the evidence combination rule of the Dempster-Shafer (D-S) theory. The approach has been applied to engineering design selection, organizational self-assessment, safety and risk assessment, and supplier assessment. In this paper, the fundamental features of the ER approach are investigated. New schemes for weight normalization and basic probability assignments are proposed. The original ER approach is further developed to enhance the process of aggregating attributes with uncertainty. Utility intervals are proposed to describe the impact of ignorance on decision analysis. Several properties of the new ER approach are explored, which lay the theoretical foundation of the ER approach. A numerical example of a motorcycle evaluation problem is examined using the ER approach. Computation steps and analysis results are provided in order to demonstrate its implementation process.     

基于证据推理规则的信息融合故障诊断方法

本文针对不确定性故障特征信息的融合决策问题,给出基于证据推理(evidence reasoning,ER)规则的故障诊断方法.首先基于故障特征样本似然函数归一化的方法求取各传感器(信息源)提供的诊断证据;从传感器误差以及故障特征对各故障类型辨别能力的差异出发,给出获取诊断证据可靠性因子的方法;给出双目标优化模型训练得到诊断证据的重要性权重,最后利用ER规则融合经可靠性因子和重要性权重修正后的诊断证据,利用融合结果进行故障决策.该方法继承了Dempster-Shafer证据理论处理不确定性信息融合问题的优点,同时克服了它在实际应用中无法区分证据可靠性和重要性的不足,使得所获诊断证据更为客观、可信.最后,通过在多功能电机转子试验台上的故障诊断实验,验证了所提方法的有效性.     

基于神经网络的机械传动方案决策系统的建模与仿真

针对机械传动方案决策系统难以建立准确数学模型的特点,提出了利用BP及BP改进神经网络的非线性映射能力,进行机械传动方案决策系统模型辨识建模的新方法.介于网络的输入输出数据难以精确描述的特点,运用模糊理论对其进行了处理.通过仿真实验验证了利用BP及BP改进网络建立其决策模型的可行性,表明了BP改进算法收敛速度更快,模型更准确,从而为机械系统传动方案的决策提供了一种更为有效的推理模型.