基于无监督学习的源数估计及盲分离算法

针对目前欠定盲分离问题中源数未知,采取"两步法"进行分离源信号.在第一步聚类算法中,一般都假设源信号个数已知,即事先给定聚类数目,这类算法成功与否依赖于源信号个数的先验知识.为了有效解决这个问题,提出了一种新的基于模糊聚类分析的无监督学习算法,它利用观测信号之间的相似关系来确定模糊相似矩阵进行迭代算法,不但可以精确估计源信号个数,同时也能获得对混叠矩阵的精确估计.该方法进一步完善了"两步法",仿真结果表明了算法的有效性及优异性能.     

基于非线性内电位构造函数的抗伪状态联想记忆

针对传统Hopfield神经网络在联想记忆中存在的大量伪状态,设计出一种以非线性内电位函数构造为核心的抗伪状态方案,以取代传统的Hebb规则及其外积法,建立了一种抗伪状态的联想记忆网络(ASS-SSNN),并通过吸引域的变化研究抗伪状态方案的正确性和有效性.仿真结果表明:ASS-SSNN比采用传统Hebb规则及其外积法的Hopfield联想记忆网络(HNN)在容错性上更优,记忆状态的吸引域明显增大,存储容量也有所增加.     

基于LMKL和OC-ELM的航空电子部件故障检测方法

针对航空电子部件故障样本获取困难以及检测准确率不高的问题,提出基于局部多核学习(localized multiple kernel learning, LMKL)和一类超限学习机(one-class extreme learning machine, OC-ELM)的故障检测方法。仅运用正常状态的小样本数据,给出了LMK-OC-ELM的数学表达形式,并在不同的门模型下推导了LMK-OC-ELM中局部核权重的优化方法;在获取局部核权重的基础上,定义了离线故障检测所需的统计检验量与阈值,以便工程实现。将所提方法应用于某型接收机,结果表明,在训练时间可控的前提下,与4种常见的一类分类(one-class classification, OCC)算法相比,所提方法可均衡地提高召回率、查准率和特异度,以LMK-OC-ELM-sig为代表,其在F1、曲线下方面积(area under curve, AUC)、G-mean和准确率4个指标上,比最近提出的局部多核异常检测(localized multiple kernel anomaly detection, LMKAD)方法分别提高了1.60%、1.57%、1.53%和2.23%。     

基于中继选择的星地协作传输系统性能分析

由于星地混合协作网络（hybrid satellite-terrestrial cooperative network,HSTCN）能显著提升卫星通信的服务质量,近年来受到了国内外学者的广泛关注,分析了基于中继选择的星地协作传输系统的性能。首先,在卫星和地面用户之间存在直达链路,并且所有中继节点在采用门限判断译码转发协议辅助卫星通信的情况下,得到用户端经过最大比合并后的输出信噪比（signal-to-noise ratio,SNR）表达式。其次,针对卫星信道服从阴影莱斯分布,而地面信道服从Nakagami-m分布,推导出随机选择和最佳选择两种中继选择策略下系统的中断概率闭合表达式。接着,进一步得到高信噪比条件下中断概率的近似表达式,为分析各种参数对系统性能的影响提供了更加便捷的方法。最后,计算机仿真不仅验证了理论推导的正确性,而且定量分析了协作传输带来的好处,从而为HSTCN技术的实际应用提供理论依据。     

Credit Scoring Based on the Set-Valued Identification Method

Credit scoring is one of the key problems in financial risk managements. This paper studies the credit scoring problem based on the set-valued identification method, which is used to explain the relation between the individual attribute vectors and classification for the credit worthy and credit worthless lenders. In particular, system parameters are estimated by the set-valued identification algorithm based on a given recognition criteria. In order to illustrate the efficiency of the proposed method, practical experiments are conducted for credit card applicants of Australia and credit card holders from Taiwan, respectively. The empirical results show that the set-valued model has a higher prediction accuracy on both small and large numbers of data set compared with logistic regression model. Furthermore, parameters estimated by the set-valued identification method are more stable,which provide a meaningful and logical explanation for extracting factors that influence the borrowers' credit scorings.     

保持轮廓清晰的有理-线性彩色图像内插放大方法

提出了一种有理-线性的自适应彩色图像向量插值方法。该方法是将数字图像的每一个象素看作是平面域的关于RGB三原色的一个向量,首先结合图像的局部和边缘特征对图像进行N×N块划分,根据图像的块特征构造二元向量Newton-Thiele型有理插值曲面或Newton-Newton插值曲面,然后对插值曲面进行采样以实现图像的缩放。采用文中的算法得到的放大图像,不仅可保持轮廓的清晰,而且可得到较好的彩色图像视觉效果。最后试验证明,该方法比常规方法产生的放大图像质量高。     

中国证券市场反向策略研究及其短周期性

证券市场中的反向策略是指购买价格被低估的"价值股票"和抛售价格被高估的"热门股票".国外大量实证研究表明,采取反向策略能获得较高的收益,而我国目前的实证结果争议较大.采用不同于以往大部分国内学者的实证方法,根据股票的价格比例对资产组合进行分类,对我国证券市场上反向策略的有效性进行了实证研究.结果表明,中国证券市场存在反应过度,反向策略行之有效,从而也说明我国证券市场的非有效性.另外,作为新兴市场,中国证券市场和西方证券市场的反向策略存在差异,主要表现在中国股市的反向周期短于西方发达国家,这可能与中国股市的高换手率有关.     

基于多准则神经网络与分类回归树的电信行业异动客户识别系统

以电信行业为应用对象,建立了一种基于多准则神经网络(MCNN)与分类回归树(CART)的的异动客户识别系统.该系统首先用多准则神经网络对客户属性进行约简,然后构造用于识别异动客户的分类回归树.通过对浙江省电信系统的大客户数据的实际验证,结果表明该系统具有较好的鲁棒性和有效性.     

The Differentiated Effect of Administrative Policy in China's Housing Market——Based on the Heterogeneous Households Multi-Agent Model

Since 2010, Chinese government has introduced a series of administrative policies to limit speculation in the housing market to stabilize price fluctuations and keep the housing market in a healthy state of development. In order to investigate whether administrative policy can play its due role, this paper constructs a comprehensive bottom-up housing market heterogeneous households multiagent model(HHMAM) to undertake research on the differentiated effect of administrative policy in different cities. The empirical studies find that: 1) Administrative policy that increases interest rates will cause housing prices to continue to decline in the long term, but they will resume a rising trend after reaching the lowest point; 2) If the government cancels a property-purchasing limitation, housing prices will continue to rise; and 3) investors tend to invest in 1~(st)-tier cities due to the high demand and greater likelihood of appreciation in these cities.     

Cooperative Mechanism of Supply Chain Under Asymmetric Information

Abstract: The cooperative mechanism is one main issue in the decentralized supply chain system, especially in an asym-metric information structure. We analyze the non-cooperative game behavior of a 2-echelon distribution supply chain,compare the results with the system optimal solution, and give the supplier dominated cooperative mechanisms. We alsoanalyze the validity of our contract under the asymmetric retailers' holding cost information and give some useful conclu-sions.