基于RBF网络的中国信贷规模稳健预测

将M-估计稳健损失函数与基于统计贡献度的动态确定核函数方法相结合, 提出一种有效的非参数RBF预测模型, 该方法克服了稳健性缺失问题, 在估计参数的同时动态确定最佳网络结构, 并且在学习中自动消除噪声和异常点的影响, 加快了网络的学习和收敛速度. 利用中国月度信贷数据进行实证分析表明, 本文模型与基准模型相比具有最好的预测稳健性和准确性, 对于提高货币政策有效性和前瞻性具有很好的应用价值.     

Radial acceleration estimation of multiple high maneuvering targets

The acceleration of a high maneuvering target in signal processing is helpful to enhance the performance of the tracker and facilitate the classification of targets. At present, most of the research on acceleration estimation is carried out in cases of a single target with time-frequency analysis methods such as fractional Fourier transform （FRFT）, Hough-ambiguity transform （HAT）, and Wigner-Vil e distribution （WVD）, which need to satisfy enough time duration and sampling theorem. Only one reference proposed a method of acceleration estimation for multiple targets based on modified polynomial phase transform （MPPT） in the lin-ear frequency modulation （LFM） continuous-wave （CW） radar. The method of acceleration estimation for multiple targets in the pulse Doppler （PD） radar has not been reported so far. Compressive sensing （CS） has the advantage of sampling at a low rate and short duration without sacrificing estimation performance. There-fore, this paper proposes a new method of acceleration estimation for multiple maneuvering targets with the unknown number based on CS with pulse Doppler signals. Simulation results validate the effectiveness of the proposed method under several conditions with different duration, measurement numbers, signal to noise ra-tios （SNR）, and regularization parameters, respectively. Simulation results also show that the performance of the proposed method is superior to that of FRFT and HAT in the condition of multiple targets.     

Image separation using wavelet-complex shearlet dictionary

This paper presents a new method for image separation through employing a combined dictionary consisting of wavelets and complex shearlets. Because the combined dictionary sparsely represents points and curvilinear singularities respectively, the image can be decomposed into pointlike and curvelike parts as accurate as possible. The proposed method based on the geo- metric separation theory introduced by Donoho in 2005 shows that accurate geometric separation of the morphologically distinct fea- tures of points and curves can be achieved by l1 minimization. The experimental results show that the proposed method can not only be effective but also greatly reduce the computing time.     

Polynomial-rooting based fourth-order MUSIC for direction-of-arrival estimation of noncircular signals

A polynomial-rooting based fourth-order cumulant algorithm is presented for direction-of-arrival（DOA） estimation of second-order fully noncircular source signals, using a uniform linear array（ULA）. This algorithm inherits all merits of its spectralsearching counterpart except for the applicability to arbitrary array geometry, while reducing considerably the computation cost.Simulation results show that the proposed algorithm outperforms the previously developed closed-form second-order noncircular ESPRIT method, in terms of processing capacity and DOA estimation accuracy, especially in the presence of spatially colored noise.     

软件可靠性关键参数判定的矩估计方法

已有的基于结构的软件可靠性研究中大多假设建模参数为已知常量,导致模型忽略了参数估计的统计特征对于系统可靠性的影响.针对该问题,提出一种基于矩估计的软件可靠性关键参数判定方法,该方法通过分析基于结构软件可靠性模型的特点,建立各建模参数的矩估计量与系统可靠性预测结果之间的关系,量化各参数对于系统可靠性的影响,判定关键参数.为系统可靠性控制、系统结构优化及软件测试资源最优化分配等问题的解决提供了理论支持,通过对两个实际软件系统的分析验证了方法的有效性.     

基于Bayes估计的有限元模型修正

讨论了基于Bayes方法来修正有限元动力模型问题 ,给出了利用Bayes方法估计参数的公式 .根据Bayes估计结果 ,应用矩阵逆特征值理论 ,提出了有限元模型修正方法 .最后用一个数值试验说明了该方法的有效性 .     

国家重点基础研究发展计划、国家重大科学研究计划2010年重要支持方向

国家重点基础研究发展计划是以国家重大需求为导向,对我国未来发展和科学技术进步具有战略性、前瞻性、全局性和带动性的基础研究发展计划。国家重大科学研究计划是《国家中长期科学和技术发展规划纲要(2006—2020年)》(以下简称《规划纲要》)部署的、引领未来发展、对科学和技术发展有很强带动作用的基础研究发展计划。围绕落实《规划纲要》,科技部2010年将继续部署国家重点基础研究发展计划和国家重大科学研究计划项目。科技部于2010年1月19日发布了2010年国家重点基础研究发展计划、国家重大科学研究计划2010年重要支持方向。此外,为贯彻落实《规划纲要》,应对气候变化自主行动、履约和国际谈判提供科技支撑,科技部2010年应急启动全球变化研究国家重大科学研究计划,针对全球变化研究中的关键科学问题,开展基础性、战略性、前瞻性研究,全面提升我国全球变化研究的竞争力。科技部于2010年2月24日发布了全球变化研究国家重大科学研究计划重要支持方向。     

用于非粘性液体检测的Lamb波传感器阵列

在负载液体的压电薄板上可通过叉指换能器(IDT)激发出Lamb波, Lamb波的相速度与液体的特性参数有关. 基于该原理, 引入有效介电常数的概念研究了Lamb波在液体传感领域的潜在应用, 并针对理想非粘性液体的检测设计了同时并行敏感液体密度和介电常数的传感器阵列. 以压电材料铌酸锂(LiNbO₃)为例, 为改进灵敏度及机电耦合系数, 通过数值计算得到了传感器阵列的板厚、切向等优化结果.     

基于深度学习的平面位姿估计算法

为了在未知物体三维模型的情况下使用深度学习进行平面位姿估计,采用编码器-解码器网络,从单个RGB图像中检测平面实例分割及法线信息,并利用这些信息进行位姿解算,获得每个平面的实时位姿。实验结果显示,平面召回率为0.625,平面法线召回率为0.414,实时性为18.5 f/s,验证了算法的可行性。     

Low complexity method for spreading sequence estimation of DSSS signal in non-cooperative communication systems

It is a necessary step to estimate the spreading sequence of direct sequence spread spectrum (DSSS) signal for blind despreading and demodulation in non-cooperative communications. Two innovative and effective detection statistics axe proposed to implement the synchronization and spreading sequence estimation procedure. The proposed algorithm also has a low computational complexity with only linear additions and modifications. Theoretical analysis and simulation results show that the algorithm performs quite well in low SNR environment, and is much better than all the existing typical algorithms with a comprehensive consideration both in performance and computational complexity.