LPNN-based approach for LASSO problem via a sequence of regularized minimizations

In compressive sampling theory, the least absolute shrinkage and selection operator (LASSO) is a representative problem. Nevertheless, the non-differentiable constraint impedes the use of Lagrange programming neural networks (LPNNs). We present in this article the -LPNN model, a novel algorithm that tackles the LASSO minimization together with the underlying theory support. First, we design a sequence of smooth constrained optimization problems, by introducing a convenient differentiable approximation to the non-differentiable -norm constraint. Next, we prove that the optimal solutions of the regularized intermediate problems converge to the optimal sparse signal for the LASSO. Then, for every regularized problem from the sequence, the -LPNN dynamic model is derived, and the asymptotic stability of its equilibrium state is established as well. Finally, numerical simulations are carried out to compare the performance of the proposed -LPNN algorithm with both the LASSO-LPNN model and a standard digital method.     

基于分段加权的反向稀疏跟踪算法研究

为提高稀疏表示跟踪模型性能，提出一种分段加权的反向稀疏跟踪算法，将跟踪问题转化为在贝叶斯框架下寻找概率最高的候选对象问题，构造不同的分段权重函数来分别度量候选目标与正负模板的判别特征系数。通过池化来降低跟踪结果的不确定性干扰，选择正负模板加权系数差值最大的候选表示作为跟踪结果。实验表明，在光照变化、遮挡、快速运动、运动模糊情况下，所提出的算法可以确保跟踪结果的准确性和鲁棒性。     

多共振分量融合CNN的行星齿轮箱故障诊断

针对行星齿轮箱中各部件所激起的振动成分混叠、早期故障特征经常被较强的各级齿轮谐波成分以及环境噪声所湮没的问题，提出一种多共振分量融合卷积神经网络（multi-resonance component fusion based convolutional neural network，简称MRCF-CNN）的行星齿轮箱故障诊断方法。首先，对振动信号进行共振稀疏分解，得到包含齿轮谐波成分的高共振分量和可能包含轴承故障冲击成分的低共振分量；其次，构建多共振分量融合卷积神经网络，将得到的高、低共振分量和原始振动信号进行自适应的特征级融合，通过有监督的方式训练模型并进行行星齿轮箱故障诊断。对行星齿轮箱实验数据的分析结果表明，该方法能够有效分类行星齿轮箱中滚动轴承和齿轮的故障，成功对行星齿轮箱故障进行诊断，同时能够进一步增强卷积神经网络对振动信号所蕴含的故障信息的辨识能力。     

Affine hull based target representation for visual tracking

Handling appearance variations is a challenging issue in visual tracking. Existing appearance models are usually built upon a linear combination of templates. With such kind of representation, accurate visual tracking is not desirable when heavy appearance variations are in presence. Under the framework of particle filtering, we propose a novel target representation for tracking. Namely, the target candidates are represented by affine combinations of a template set, which leads to better capability in describing unseen target appearances. Additionally, in order to adapt this representation to dynamic contexts across a video sequence, a novel template update scheme is presented. Different from conventional approaches, the scheme considers both the importance of one template to a target candidate in the current frame and the recentness of the template that is kept in the template set. Comprehensive experiments show that the proposed algorithm achieves superior performances in comparison with state-of-the-art works.     

基于双向位图的CSR大规模图存储优化

大数据时代,Graph500是评测超级计算机处理数据密集型应用能力的重要工具,E级验证系统的图遍历处理能力主要受限于内存空间和访存带宽,尤其是内存空间利用率直接决定了图的测试规模和测试性能.针对天河E级验证系统小内存特征,提出了基于双向位图的大规模图数据压缩存储方法(bidirectional-bitmap based CSR,Bi-CSR),Bi-CSR在CSR矩阵压缩的基础上引入行方向位图和列方向位图协同完成稀疏矩阵压缩存储,行方向位图主要负责行方向位图的压缩存储与索引,列方向位图除了进一步压缩图存储空间,还负责为顶点遍历向量并行优化提供加速空间.Bi-CSR大幅度减少了稀疏矩阵存储空间.面向天河E级验证系统,当图输入规模为237时,Graph500的图存储空间节约效率接近70%,全系统稳定测试性能为2.131E+12TEPS,性能最大加速比超过100倍.     

结构化加权稀疏低秩恢复算法在人脸识别中的应用

针对训练样本或测试样本存在污损的情况，提出一种结构化加权稀疏低秩恢复算法（structured and weighted-sparse low rank representation，SWLRR）。SWLRR对低秩表示进行加权稀疏约束和结构化约束，使得低秩表示系数更加趋近于块对角结构，进而可获得具有判别性的低秩表示。SWLRR将训练样本恢复成干净训练样本后，再根据原始训练样本和恢复后的训练样本学习到低秩投影矩阵，把测试样本投影到相应的低秩子空间，即可有效地去除测试样本中的污损部分。在几个人脸数据库上的实验结果验证了SWLRR在不同情况下的有效性和鲁棒性。     

Off‐grid DoA estimation method of wideband signals

In this paper, an off‐grid direction of arrival (DoA) estimation method is proposed for wideband signals. This method is based on the sparse representation (SR) of the array covariance matrix. Similar to the time domain DoA estimation methods, the correlation function of the sources was assumed to be the same and known. A new measurement vector is obtained using the lower‐left triangular elements of the covariance matrix. The DoAs are estimated by quantizing the entire range of continuous angle space into discrete grid points. However, the exact DoAs may be located between two grid points; therefore, this estimation has errors. The accuracy of DoA estimation is improved by the minimization of the difference between the new measurement vector and its estimated values. Simulation results revealed that the proposed method can enhance the DoA estimation accuracy of wideband signals.     

基于冗余字典的联合稀疏同步迭代硬阈值算法

为了改进基于压缩感知(CS)的欠Nyquist采样系统在冗余字典条件下信号重构的效果,研究了基于ε-闭包的分块联合稀疏模型的同步迭代硬阈值(SIHT)算法。分析了采样系统基于多测量向量(MMV)的CS合成模型,提出了ε-闭包的分块相干性和约束等距特性(RIP)概念;在迭代过程中根据冗余字典分块相干性,对更新支撑集进行优选从而完成算法改进;给出了迭代收敛常数,并分析了改进型算法的收敛特性。仿真实验结果表明,相比传统算法,改进型算法在采样系统足够的通道数条件下重构成功率可达到100%,噪声抑制能力能够提高7 dB~9 dB,总运算时间可以降低至少37.9%,信号重构收敛速度更快。     

基于稀疏表示的增维叠前地震反演方法

常规的叠前和叠后反演技术难以满足刻画薄储层的需求。由于道集叠加而产生的干涉效应对地震信号所蕴含的真实信息造成不同程度的弱化、畸变甚至湮灭，因此为有效地保存真实信息，应尽量避免叠加。另外，叠前道集的地震属性在随炮检距变化之外还受频率因素的影响，而以高分辨率谱分解为基础展开的FAVO反演方法能够提升流体识别能力。为此，以不叠加为核心思想，提出基于稀疏表示的增维叠前地震反演方法。首先，从叠前道集出发，对包含目的层段的角度内道集进行筛选和提取，依托信号的稀疏表示理论对抽选的单角度地震数据开展高分辨率时频分解；其次，基于贝叶斯理论直接建立波阻抗与地震数据的映射关系，利用非线性最优化算法对初始模型进行扰动；最后，以前一个频率反演的结果作为后一个频率反演的约束，逐级进行反演，最终得到不同角度的高分辨率增维反演结果。以A研究区为例，在有效角度范围内对单角度地震数据分频约束反演，拓宽了数据维度，提升了反演精度，增强了对薄储层的描述能力。     

采用多波前法求解大型结构方程组

多波前法算法可以求解十万个自由度以上的大型结构方程组,其求解速度和存储效率都优于传统算法,并可统一处理病态矩阵。基于大型工程的数值模拟验证了多波前法的稳定与高效。该方法已应用在深圳市广厦软件有限公司的"建筑结构通用分析与设计软件GSSAP"中。