基于隐Markov树的设备状态综合诊断模型

为了充分利用小波系数之间的统计相依性以更有效地诊断设备状态,提出了一种基于隐Markov树(HMT)的综合诊断模型。首先通过主成分分析将来自多个传感器的信号转换为主成分,求出各主成分对应的频谱,然后通过比较对已训练的各HMT模型的适应度,运用Bayes决策融合法则得到设备状态综合诊断决策。为了克服HMT模型存在的计算溢出困难,采用尺度变换对EM算法进行了改进。通过两个实例验证了该综合诊断模型具有较高的诊断准确率。     

基于遗传算法的RLV再入轨迹优化设计

飞行器再入轨迹优化是一类最优控制问题。传统的优化方法存在初始值敏感问题。利用小生境技术和精英方法改进适应值共享拥挤遗传算法,并将其用于RLV再入轨迹优化设计。以终端时间固定的空间最小控制能量再入轨迹和终端时间自由的平面最小热载再入轨迹为例,详细讨论了遗传算法用于再入轨迹优化设计所需要解决的一些关键问题。仿真结果表明提出的方法能够较快地搜索到全局最优解,对初始猜测值不敏感,能够方便用于RLV的再入轨迹方案选择和优化设计。     

改进的聚类算法及在复杂大群体决策中的应用

针对现有FCM聚类算法中存在的局部极值和伸缩性较差等问题,提出了基于全部最小连通支配集算法(minimum connected donating set algorithm,MCDSA)的改进的聚类算法(minimum fuggy C-means,MF-CM)。用改进的聚类算法MFCM辅助复杂大群体决策(complex huge group-decision,CHGDS),定义了群体偏好矢量和群体一致性指标,提供了一种新的解决CHGDS中群体决策的理论和方法,并通过实验证明了该方法的有效性和稳定性。还提供了基于属性加权进化的群体一致性决策机制新思路。     

基于对角递归神经网络盲均衡算法的研究

提出了一种基于对角递归神经网络的盲均衡算法。利用对角递归神经网络结构简单、计算量少的优点,结合传统的恒模盲均衡算法定义了代价函数,用最速梯度下降法推导出了其算法迭代公式。计算机仿真表明,该算法收敛速度较快,误码率较小。     

Research and Application on Web Information Retrieval Based on Improved FP-Growth Algorithm

0 IntroductionThe useful texts information in the Web is analyzed bythe method of Web information retrieval according tothe requirement of the user . To get the information of citingtexts and relative key words fromthe Web,we can searchforlots of correlativeinformation data storedinthe Web databaseby Webinformationretrieval .This paper appliedthe method-ology of data mining to Web information retrieval since thepurpose of data miningisto discover thefrequent patternfromthe Web.Alot of studi…     

A Parallel Genetic Simulated Annealing Hybrid Algorithm for Task Scheduling

0 IntroductionGrid computingis a hot topic inthe current internet research,and a developing direction of the parallel and distributedprocess[1 ,2].Since the task scheduling in grid computing faces aNP-hard problem[3];it has drawn attention from many scholarsand become the focusinthe field of the current grid computing re-search.In recent years , two global random and opti mal algorithmhave been widelystudiedandappliedinthefield of the gridcompu-ting research: GA(Genetic Algorithm) and SA( …     

Dynamically Reconfigurable Encryption System of the AES

Reconfigurable computing has grown to become an important and large field of research, it offers advantages over traditional hardware and software implementations of computational algorithms. The Advanced Encryption Standard （AES） algorithm is widely applied in government department and commerce. This paper analyzed the AES algorithms with different cipher keys, adopted a novel key scheduler that generated the round key real-time, proposed a dynamically reconfigurable encryption system which supported the AES algorithm with different cipher keys, and designed the architecture of the reconfigurable system. The dynamically reconfigurable AES system had been realized on FPGA. The result proves that the reconfigurable AES system is flexible, lower cost and high security level.     

Specification and Verification for Semi-Structured Data

Tree logic, inherited from ambient logic, is introduced as the formal foundation of related programming language and type systems, In this paper, we introduce recursion into such logic system, which can describe the tree data more dearly and concisely. By making a distinction between proposition and predicate, a concise semantics interpretation for our modal logic is given. We also develop a model checking algorithm for the logic without △ operator. The correctness of the algorithm is shown. Such work can be seen as the basis of the semi-structured data processing language and more flexible type system.     

新的ICA算法实现成组fMRI信号盲分离

独立成分分析(ICA)方法已被成功地用于处理功能磁共振成像(fMRI)信号,但主要是用于处理单个被试的fMRI信号,对于多个被试的情况却很少考虑.为此利用一种扩展的ICA方法--Group ICA来处理多个被试的fMRI信号,结果表明这种方法在保证结果准确性的前提下,可以大大减少计算量,快速获得统计结果.计算中应用的是NewFP算法,统计结果表明这种算法在估计激活的时间动力学准确性上优于FastICA算法.     

神经网络在项目评估中的应用

项目评估在市场机遇获得、成本控制和竞争力判别等方面决定着企业的成功与否.通过对影响项目价值的属性、因素、显性指标以及它们之间关系的研究,采用模拟生物神经元基本功能的神经网络技术建立项目价值模型,改进了传统的BP算法,对样本进行监督训练的反向传播算法得到权重因子,从而确定指定项目的价值,辅助决策者进行项目评估决策.实际应用证明了算法的有效性和模型的实用性.