Adaptive discretization for signal detection in statistical inverse problems

We discuss statistical tests in inverse problems when the original equation is replaced by a discretized one, i.e. a linear system of equations. Previous studies revealed that using the discretization level as regularizing procedure is possible, but its application is limited unless discretization is restricted to the singular value decomposition, see C. Marteau and P. Mathé, General regularization schemes for signal detection in inverse problems, 2013. General linear regularization may circumvent this, and we propose a regularization of the discretized equations. The discretization level may be chosen adaptively, which may save computational budget. This results in tests which are known to yield the optimal separation rate up to some constant in many cases.     

Meshless acoustic analysis using a weakly singular Burton-Miller boundary integral formulation

The Burton-Miller boundary integral formulation is solved by a complex variable boundary element-free method (CVBEFM) for the boundary-only meshless analysis of acoustic problems with arbitrary wavenumbers. To regularize both strongly singular and hypersingular integrals and to avoid the computation of the solid angle and its normal derivative, a weakly singular Burton-Miller formulation is derived by considering the normal derivative of the solid angle and adopting the singularity subtraction procedures. To facilitate the implementation of the CVBEFM and the approximation of gradients of the boundary variables, a stabilized complex variable moving least-square approximation is selected in the meshless discretization procedure. The results show the accuracy and efficiency of the present CVBEFM and reveal that the method can produce satisfactory results for all wavenumbers, even for extremely large wavenumbers such as k = 10 000.     

基于L1正则化的随机梁式结构静力损伤识别方法

提出了一种结合摄动法和L1正则化方法的随机梁式结构静力损伤识别方法。考虑初始模型误差和测量误差的影响,建立了关于随机损伤指数的控制方程,并将摄动法和L1正则化方法相结合,对随机损伤指数的控制方程进行求解,进而从概率的角度对结构的损伤进行识别。损伤试验结果表明,和传统的最小二乘求解法相比,本文方法能够更为准确地识别多处局部损伤的位置及大小,对实际结构损伤检测具有较好的参考价值。     

On Regularization of a Source Identification Problem in a Parabolic PDE and its Finite Dimensional Analysis

We consider the inverse problem of identifying a general source term, which is a function of both time variable and the spatial variable, in a parabolic PDE from the knowledge of boundary measurements of the solution on some portion of the lateral boundary. We transform this inverse problem into a problem of solving a compact linear operator equation. For the regularization of the operator equation with noisy data, we employ the standard Tikhonov regularization, and its finite dimensional realization is done using a discretization procedure involving the space $L^2(0,\tau;L^2(Ω))$. For illustrating the specification of an a priori source condition, we have explicitly obtained the range space of the adjoint of the operator involved in the operator equation.     

HMD-Net: A Vehicle Hazmat Marker Detection Benchmark

Vehicles carrying hazardous material (hazmat) are severe threats to the safety of highway transportation, and a model that can automatically recognize hazmat markers installed or attached on vehicles is essential for intelligent management systems. However, there is still no public dataset for benchmarking the task of hazmat marker detection. To this end, this paper releases a large-scale vehicle hazmat marker dataset named VisInt-VHM, which includes 10,000 images with a total of 20,023 hazmat markers captured under different environmental conditions from a real-world highway. Meanwhile, we provide an compact hazmat marker detection network named HMD-Net, which utilizes a revised lightweight backbone and is further compressed by channel pruning. As a consequence, the trained-model can be efficiently deployed on a resource-restricted edge device. Experimental results demonstrate that compared with some established methods such as YOLOv3, YOLOv4, their lightweight versions and popular lightweight models, HMD-Net can achieve a better trade-off between the detection accuracy and the inference speed.     

接触力优化设计的熵正则化方法

包含单侧接触的机械和结构系统中,接触面间的局部接触力很大时,容易导致结构局部磨损和破坏。利用结构优化的手段极小化接触面上的最大接触力是调整接触面上接触力分布的一个有效途径。然而在数学上,目标函数的不可微和约束中的接触条件导致该结构优化问题属于非光滑和非凸优化问题,因而常用优化算法的优化结果是不可靠的。为此,本文提出接触力优化的一个熵正则化模型,并利用基于正则化过程的序列二次结构优化算法求解。所提的方法在数值上避免了处理不可微问题,而且正则化参数的调整还有助于获得更小的局部最优解。以离散结构为例的数值算例验证了本文的模型和算法。     

求解参数识别反问题的同伦正则化方法

提出了求解参数识别反问题的同伦正则化方法,给出了相应的收敛性定理.数值结果表明该方法是一种快速的大范围收敛方法.     

Differentially expressed genes selection via Laplacian regularized low-rank representation method

With the rapid development of DNA microarray technology and next-generation technology, a large number of genomic data were generated. So how to extract more differentially expressed genes from genomic data has become a matter of urgency. Because Low-Rank Representation (LRR) has the high performance in studying low-dimensional subspace structures, it has attracted a chunk of attention in recent years. However, it does not take into consideration the intrinsic geometric structures in data.In this paper, a new method named Laplacian regularized Low-Rank Representation (LLRR) has been proposed and applied on genomic data, which introduces graph regularization into LRR. By taking full advantages of the graph regularization, LLRR method can capture the intrinsic non-linear geometric information among the data. The LLRR method can decomposes the observation matrix of genomic data into a low rank matrix and a sparse matrix through solving an optimization problem. Because the significant genes can be considered as sparse signals, the differentially expressed genes are viewed as the sparse perturbation signals. Therefore, the differentially expressed genes can be selected according to the sparse matrix. Finally, we use the GO tool to analyze the selected genes and compare the P-values with other methods.The results on the simulation data and two real genomic data illustrate that this method outperforms some other methods: in differentially expressed gene selection.     

Depth-adaptive regularized reconstruction for reflection diffuse optical tomography

We study reflection diffuse optical tomography using two-dimensional (2D) continuous-wave source-detector arrays on the surface of semi-infinite medium, aiming at imaging the perfusion and the hemoglobin oxygen saturation variation of human cerebral cortex with brain activation. We had previously formulated the inverse problem with Moore-Penrose inversion. When we use simple regularization in this inverse problem, the reconstruction sensitivity decreases markedly with the depth so that the signal in the deep range may be masked by an unwanted signal in the shallow range. In this paper, we propose a depth-adaptive regularized reconstruction, in which we assign a smaller regularization parameter with the depth. We demonstrate improvement of the three-dimensional (3D) reconstruction uniformity using the proposed scheme.