一种基于小生境遗传算法的电网信息安全风险评估模型

信息安全保障对于电力信息物理系统安全稳定运行至关重要,其关键在于对电力信息物理系统进行全方位实时监控,并对采集到的海量监测数据进行分析以做出准确的安全风险评估结果。作为用于模式分类的进化算法,基因表达式编程(gene expression programming,GEP)算法由于其可以执行全局搜索而受到广泛关注,但其在高维度数据集下的运算极为耗时。针对上述问题,提出了一种基于小生境提高样本多样性的改进基因表达式编程算法用于电网信息安全风险评估,该算法首先利用粗糙集的思想,通过分辨函数求解最优属性对数据样本进行约简,再利用小生境模型提高约简样本个体的多样性以加快GEP算法运算的收敛速度,进而通过遗传算法实现全局搜索并得到安全风险等级评估结果。仿真实验表明,与传统的安全风险评估算法相比,提出的改进GEP算法具有较高的属性约简率和全局收敛率,可以快速实现海量监测数据下的电网信息安全风险评估。     

凸轮机构的遗传优化设计

针对标准遗传算法在解最优化问题时存在早熟收敛和后期收敛缓慢的现象，提出了一种基于小生境技术的改进遗传算法，然后结合凸轮机构的优化6设计原理，建立凸轮机构的遗传优化设计模型，并给出其应用实例和优化结果。     

IT for niche companies: is an ERP system the solution?

Abstract.  Niche companies are per definition idiosyncratic. They survive in a competitive world by mastering a small market niche, providing what their customers need. This often requires a flexible organization, and the ability to customize products. To be more efficient, many of these companies rely on extensive use of IT, often by installing general Enterprise Resource Planning (ERP) systems. These systems have grown from isolated systems that handle planning based on incoming orders and the component structure of the various products, to systems with ambitions to embrace the total functioning of the company including vendor and customer relation management. In this paper, we present four case studies. One company is a part of a large enterprise, but performs niche functions within this enterprise. The other three are small- or medium-sized enterprises. Each of these performs in small niche markets. Common to all is the fact that they encounter problems with the utilization of their ERP systems. The major problem seems to be that the ERP system has an inherent business model that may not conform to the needs of the company. Without a good understanding of the underlying models and the constraints under which the fundamental algorithms operate, it is difficult to use these systems correctly. Even excellent systems may give bad results if they are applied to situations where they are not suited. Further, the monolithic structure of an ERP system, with a rather complicated parameter setting, is often insufficient to mould the system to the needs of a niche company. We discuss these problems based on our four case studies, and offer alternative approaches that may be considered.     

小生境遗传算法AVOA反演

采用具有水平对称轴的横向各向同性(HTI)模型研究裂缝性油气藏,P波振幅随炮检距和方位角变化(AVOA),表现出较强的方位各向异性特征。本文基于Ruger提出的HTI介质模型的P波反射系数公式,利用由基准测线和另外两条与之呈45°、90°测线的P波反射系数的相对差异估算裂缝方位;同时引入小生境遗传算法,对基准测线和与之呈90°测线的P波反射系数差进行反演,得到了较高精度的纵、横波波速比和Thomsen各向异性参数。     

一种改进的共同进化遗传算法及其应用

将共同进化遗传算法应用于临床营养决策优化中,虽然取得了一定成果,但算法中仍存在早熟现象、局部搜索能量弱等问题,为此提出一种改进的共同进化模型。算法使用新的子群体划分方法和交叉算子,引入精英保留策略、信息交换原则和小生境技术。通过对糖尿病营养治疗的仿真表明,改进后的算法在保证群体多样性和避免早熟现象的前提下,提高了算法的局部和全局寻优能力,且膳食配方中营养素达标率符合要求,满足专家建议的糖尿病饮食需求。     

Research on Adaptability of Construction Project from Point of Ecology

Concept of zoology energy is advanced, zoology energy development trend line and zoology energy development curve of construction project are traced out in this paper. From the point of construction project itself, its adaptability can be shown in its own strong zoology energy, dominant in zoology energy sequence of its engineering population, great niche breath and less superpostion (overlap) in its engineering community. Adaptability of construction project is researched from the angle of ecology which can be taken as basic research of consistent development of construction project, and enrich evaluation system of construction project.     

用于织物疵点检测的最优Gabor滤波器设计

提出一种基于织物纹理特征的最优Gabor滤波器设计方法．分别建立了正常纹理匹配和疵点纹理匹配的Gabor滤波器优化设计模型，并采用小生境遗传算法对两种模型进行求解．通过比较和分析两种滤波器的检测结果发现，由正常纹理匹配模型得到的最优Gabor滤波器更适宜于织物疵点的识别与分割，并且其中心频率与纹理图像功率谱中能量最集中的谐波成分相一致，因而可以极大地缩短求解优化模型所花费的时间．     

基于中心定位算子的遗传算法

针对遗传算法局部搜索能力弱,求解精度不高的缺陷提出了一种中心定位算子.在进化一定代数(T)后,选择最优的若干个(N)染色体基因来计算中心定位算子,从而确定与中心定位算子相同的基因位,并且在以后的交叉、变异操作中,都不让相同的基因位参与.随着算法的进行,染色体相同的基因位逐渐全部地被确定下来.其次,通过与小生境技术的局部搜索能力算法的结合,提高了该算子的全局优化能力;最后,通过几个非常容易陷入局部最优的测试函数测试表明几乎所有的峰值都得到了理论值.     

基于生态位理论的虚拟企业BP神经网络评价模型研究

借鉴了生态学中生物生态位的概念和理论,分析了虚拟企业生态位概念的内涵.从虚拟企业生存、进化和竞争的三维生存空间角度,建立了基于"态"、"势"、"位"概念的虚拟企业生态位评价体系.构建了基于BP神经网络的虚拟企业生态位评价模型,利用MATLAB神经网络软件包进行仿真计算,并通过典型算例验证了该方法的有效性.     

基于自适应小生境混合遗传算法的说话人识别

为了解决传统高斯混合模型(Gaussian Mixture Model,GMM)对初值敏感,在实际训练中极易得到局部最优参数的问题,本文提出了一种GMM参数优化的新方法.将小生境技术与最大似然估计融入到遗传训练过程,形成了一种新的混合算法,缓解了遗传算法产生的"早熟"现象,提高了算法的局部搜索能力.采用自适应策略来控制交叉和变异算子,同时在适应度评价中融入了其他用户的区分性信息,提高了模型的分类精度,增强了GMM的泛化能力.实验表明,与传统和改进的两种方法相比,本文的方法都可以得到更优的模型参数,使得系统的识别率进一步提高.