基于遗传神经网络的全天户外场景红外热像仿真

针对红外场景仿真中使用热力学方法对物体温度场建模的不足,将BP神经网络应用到场景中物体的红外温度场建模中,并通过遗传算法优化神经网络的初始权值.通过对场景中物体的表观温度进行多次测量,得到训练样本集合.然后训练神经网络,建立由车辆和路面组成的场景的温度场模型,并根据设定的气象条件分别对白天和夜晚场景进行仿真.根据仿真结果分析,此模型能够根据所设定的气象条件较准确地实时仿真场景的红外图像.     

基于频率采样技术的FIR数字滤波器优化设计

本文提出了一种FIR数字滤波器满意优化设计方法，将滤波器过渡带样本值作为优化变量，通过设计通带、过渡带和阻带性能指标满意度函数和综合满意度函数，构造出满意优化模型，并用本文提出的新量子遗传算法搜索满意解，FIR高通和带阻数字滤波器的设计结果表明，采用满意优化方法设计的FIR滤波器的性能优于传统方法。     

设计IIR数字滤波器的遗传优化算法

提出了ＩＩＲ数字滤波器设计的遗传优化法。这是一种模拟自然遗传和达尔文进化理论的随机并行优化算法。首先，详细描述了遗传算法并给出了计算步骤，然后将遗传算法用于ＩＩＲ数字滤波器的优化设计，最后给出了模拟计算结果。     

A genetic algorithm solution to the design of slat-type shading system

This article presents a technique for the design of slat-type blinds based on the their relative light intensity distribution under a uniform light source. The technique offers a new approach using a genetic algorithm in order to evolve the design according to a set of parameters.     

基于协同矩阵分解的评分与信任联合预测

信息评分预测和信任预测是社交评价网络中的两大基本问题.为应对在提高两类基本问题预测准确性过程中遇到的评分数据与信任关系数据稀疏问题,本文提出了一种基于协同矩阵分解的信息评分与信任预测联合模型.该模型在将评分矩阵与信任关系矩阵进行协同分解时,既能保证被分解的两个矩阵分解过程共享用户潜在变量,又能兼顾两个矩阵分解过程中能够各自获得反映本领域知识相关性的表达.使用分解得到的多个相关低维潜在变量矩阵乘积即可做出评分与信任两个问题的预测.两个真实网络数据集上的实验验证了提出模型有效性和先进性.     

面向离散文本舆情分析的分聚类方案

离散文本已经成为一种占据重要地位的舆情信息表现形式,根据离散文本的特点,提出基于特征概念网的离散文本舆情信息的分聚类框架,在此基础上给出分聚类方案。在聚类算法中,运用了遗传算法的全局并行搜索能力、k—means的高效局部聚类能力和小生境的保持种群多样性抑制漂移能力;在分类算法中,先将训练文本库进行类内聚类成子类,对子类构建特征概念网以生成替代该子类的文本,再用KNN算法进行分类。最后结合舆情分析进一步提出了可用的改进方案。     

定向传感器网络中的最小化覆盖间隙和最大化网络生命时间问题的研究

本文研究了定向传感器网络中最小化覆盖间隙和最大化网络生命时间的问题.本文采用的定向感知天线模型,每个传感器有多个感应方向.在无线传感器网络中,最大化网络生命时间和最小化覆盖间隙是两个冲突的目标.为了在两者之间做出权衡,文章研究了在生命时间受约束的情况下最小化覆盖间隙问题(MCBLC)和在覆盖间隙受约束的条件下最大化网络生命时间问题(MLCBC).对于MCBLC问题,我们首先将它模型化为整数规划问题,并提出两个启发式算法(MCBLC-G和MCBLC-G-1).基于MCBLC-G(MCBLC-G-1)算法,利用二分搜索技术得到MLCBC问题的算法.最后,模拟验证了算法的性能.     

Analysis and Comparison of Location Strategies for Reducing Registration Cost in PCS Networks

In mobile environments, a personal communication service (PCS) network must keep track of the location of each mobile user in order to correctly deliver calls. A basic scheme used in the standard IS-41 and GSM protocols is to always update the location of a mobile user whenever the mobile user moves to a new location. The problem with this approach is that the cost of location update operations is very high especially when the mobile user moves frequently. In recent years, various location management strategies for reducing the location update cost have been proposed. However, the performance issue of these proposed algorithms remains to be investigated. In this paper, we develop two Markov chains to analyze and compare the performance of two promising location update strategies, i.e., the two location algorithm (TLA) and the forwarding and resetting algorithm (FRA). By utilizing the Markov chain, we are able to quickly answer what-if questions regarding the performance of PCS networks under various workload conditions and also identify conditions under which one strategy can perform better than the others. Using the cost due to location update and search operations between two successive calls to a mobile user as a performance measure, we show that when the mobile user exhibits a high degree of locality and the mobile user's call-to-mobility ratio (CMR) is low, TLA can significantly outperform both FRA and IS-41. On the other hand, when CMR is high, FRA is the winner. Furthermore, unlike TLA which may perform worse than IS-41 at high CMR values, FRA at identified optimal conditions can always perform at least as good as IS-41, suggesting that FRA over TLA should be used as a generic scheme in reducing the location cost in the PCS network to cover all possible per-user CMR values.     

基于自适应遗传算法多项式预失真技术研究

提出了一种基于自适应遗传算法的多项式预失真补偿方法,该方法的基本原理是利用模拟进化类算法对目标函数进行逼近,对待求解参数进行辨识,进而得到多项式预失真模型。该算法具有良好的收敛性、全局性和鲁棒性,且具有较高的计算效率,为解决数字预失真自适应算法提供了一个新的参考。同时将自适应遗传算法应用于功放线性化设计中并进行了仿真。仿真分析表明,采用模拟进化类的遗传算法对功放失真抑制效果明显。