Spare L1-norm-based maximum margin criterion

Maximum margin criterion (MMC) is a popular method for dimensionality reduction or feature extraction. MMC can alleviate the small size sample (SSS) problem encountered by linear discriminant analysis (LDA) and extract more discriminant vectors than LDA. However, the objective function of MMC is derived from L2-norm, which makes MMC be sensitive to noise and outliers. Besides, the basis vectors of MMC are dense, which makes it hard to explain the obtained features. To address the drawbacks of MMC, in this paper, we propose a novel sparse L1-norm-based maximum margin criterion (SMMC-L1). L1-norm rather than L2-norm is used in the objective function of SMMC-L1. Besides, L1-norm is also used as a lasso penalty to regularize the basis vectors. An iterative algorithm for solving SMMC-L1 is proposed. Experiment results on some databases show the effectiveness of the proposed SMMC-L1.     

基于L1-范数的最大间距准则

在进行线性投影降维时,由于传统的最大间距准则(Maximum Margin Criterion,MMC)算法基于L2-范数,易于受到野值(outliers)及噪声的影响.该文提出一种基于L1-范数的最大间距准则(L1-norm-based MMC,MMC-L1)降维方法,它充分利用L1-范数对野值及噪声的强鲁棒性以及最大间距准则,提出了一种快速迭代优化算法,并给出了其单调收敛到局部最优的证明.在多个图像数据库上的实验验证了该方法的鲁棒性与高效性.     

基于戴帽L1范数的双支持向量机

针对双支持向量机模型易受异常点影响导致泛化性能较低的问题,提出了一种基于戴帽L1范数的双支持向量机模型.采用带有上限值的戴帽L1范数代替L2范数来构造最优化问题,一定程度上削弱了离群点、噪音点对于两个超平面构造的影响,增强了模型的鲁棒性.另外,针对构造的新的双支持向量机模型最优化问题提出了一个简单有效的迭代算法并且在理论上证明了该算法的收敛性.在无噪以及有噪UCI数据集上的实验结果表明,与其它支持向量机模型相比,该模型有着更强的鲁棒性以及稳定性.     

Robust L1-norm two-dimensional collaborative representation-based projection for dimensionality reduction

Collaborative representation-based projection (CRP) is a well-known dimensionality reduction technique, which has been proved to have better performance than sparse representation-based projection (SRP) in the fields of recognition and computer vision. However, classical CRP is sensitive to noises and outliers since its objective function is based on L2-norm, and it will suffer from the curse of dimensionality as it is used for images processing. In this paper, a novel CRP model, named L1-norm two-dimensional collaborative representation-based projection (L1-2DCRP) and an efficient iterative algorithm to solve it are proposed. Different from conventional CRP, the optimal problem in our proposed model is a L1-norm-based maximization and the vector data is extended to matrix date. The proposed algorithm is theoretically proved to be monotonously convergent, and more robust to noises and outliers since L1-norm is used. Experimental results on CMU Multi-PIE, COIL20, FERET and ORL face databases validate the effectiveness of L1-2DCRP compared with several state-of-the-art approaches.     

L1-norm-based common spatial patterns

Common spatial patterns (CSP) is a commonly used method of spatial filtering for multichannel electroencephalogram (EEG) signals. The formulation of the CSP criterion is based on variance using L2-norm, which implies that CSP is sensitive to outliers. In this paper, we propose a robust version of CSP, called CSP-L1, by maximizing the ratio of filtered dispersion of one class to the other class, both of which are formulated by using L1-norm rather than L2-norm. The spatial filters of CSP-L1 are obtained by introducing an iterative algorithm, which is easy to implement and is theoretically justified. CSP-L1 is robust to outliers. Experiment results on a toy example and datasets of BCI competitions demonstrate the efficacy of the proposed method.     

基于L1-范数的二维线性判别分析

为了避免图像数据向量化后的维数灾难问题,以及增强对野值(outliers)及噪声的鲁棒性,该文提出一种基于L1-范数的2维线性判别分析(L1-norm-based Two-Dimensional Linear Discriminant Analysis, 2DLDA-L1)降维方法。它充分利用L1-范数对野值及噪声的强鲁棒性,并且直接在图像矩阵上进行投影降维。该文还提出一种快速迭代优化算法,并给出了其单调收敛到局部最优的证明。在多个图像数据库上的实验验证了该方法的鲁棒性与高效性。     

Residues cluster-based segmentation and outlier-detection method for large-scale phase unwrapping

2-D phase unwrapping is an important technique in many applications. However, with the growth of image scale, how to tile and splice the image effectively has become a new challenge. In this paper, the phase unwrapping problem is abstracted as solving a large-scale system of inconsistent linear equations. With the difficulties of large-scale phase unwrapping analyzed, L(0)-norm criterion is found to have potentials in efficient image tiling and splicing. Making use of the clustering characteristic of residue distribution, a tiling strategy is proposed for L(0)-norm criterion. Unfortunately, L(0)-norm is an NP-hard problem, which is very difficult to find an exact solution in a polynomial time. In order to effectively solve this problem, equations corresponding to branch cuts of L(0)-norm in the inconsistent equation system mentioned earlier are considered as outliers, and then an outlier-detection-based phase unwrapping method is proposed. Through this method, a highly accurate approximate solution to this NP-hard problem is achieved. A set of experimental results shows that the proposed approach can avoid the inconsistency between local and global phase unwrapping solutions caused by image tiling.     

应用于水下生物识别的联合范数主成分分析算法北大核心CSCD

针对F范数对离群数据较为敏感,而L1范数能降低离群数据的影响,但无法有效控制重构误差的问题,本文将L1范数与F范数同时作为目标函数的距离度量方式,提出了二维主成分分析(two-dimensional principle component analysis,2DPCA)联合算法2DPCA-F-L1,并给出了其非贪婪求解方法。该算法确保了对图像的分类能力,同时也降低了图像重构时的平均重构误差。本文将提出的2DPCA-F-L1算法在应用于水下生物图像识别时,可以抑制水下光学影像存在的噪声干扰。实验证明,该算法能够精确地识别水下生物的种类,并且在图像重构时相较于其他主成分分析(principle component analysis,PCA)算法具有更优的鲁棒性。     

一种基于L_∞范式的求解已知旋转相机重建算法

基于多幅图像序列的三维重建过程中,相机模型的坐标统一是非常重要的基础。在已知两两相机之间的相对关系的情况下,将相机模型统一至同一世界坐标系,并恢复特征点的三维坐标。本文给出了一种基于L∞范式的求解已知旋转相机重建方法。给出了一种基于L∞范式的几何结构和运动问题的新框架,并用实验证明了算法能够达到很好的性能。     

模拟电路故障诊断L_1估计及其神经网络解法

基于精确罚函数法,提出了新的求解L_1范致问题最优解的神经网络方法,它避免了Kennedy和Chua(1988)网络罚因子较大时性态变坏问题。对Bandler(1982)提出的模拟电路故障诊断L_1范数法进行了改进,将线性约束L_1问题转化为非线性约束L_1问题,并用新的神经网络方法求解,计算量小。模拟实验表明,所提神经网络方法和改进的模拟电路故障诊断L_1范数方法是可行的。     

A NEURAL-BASED NONLINEAR L_1-NORM OPTIMIZATION ALGORITHM FOR DIAGNOSIS OF NETWORKS*

Based on exact penalty function, a new neural network for solving the L1-norm optimization problem is proposed. In comparison with Kennedy and Chua's network(1988), it has better properties.Based on Bandler's fault location method(1982), a new nonlinearly constrained L1-norm problem is developed. It can be solved with less computing time through only one optimization processing. The proposed neural network can be used to solve the analog diagnosis L1 problem. The validity of the proposed neural networks and the fault location L1 method are illustrated by extensive computer simulations.     

Robust decision design using a distance criterion

An alternate formulation of the robust hypothesis testing problem is considered in which robustness is defined in terms of a maximin game with a statistical distance criterion as a payoff function. This distance criterion, which is a generalized version of signal-to-noise ratio, offers advantages over traditional error probability or risk criteria in this problem because of the greater tractability of the distance measure. Within this framework, a design procedure is developed which applies to a more general class of problems than do earlier robustness results based on risks. Furthermore, it is shown for the general case that when a decision rule exists that is robust in terms of risk, the same decision rule will be robust in terms of distance, a fact which supports the use of the latter criterion.     

混合正则化约束的湍流退化图像复原算法

针对大气湍流引起的红外图像模糊问题,提出一种基于混合正则化的模糊核估计模型。根据图像主要边缘的稀疏性,采用图像梯度的L0范数为正则化项;通过分析模糊核的特性,提出能适用于复杂模糊情况的核L0-L2范数正则化约束。复原模型的优化过程中,结合变量分裂策略和增广拉格朗日法交替估计图像和模糊核,并利用快速傅里叶变换,实现模糊核的快速、准确估计;最终根据估计的模糊核,复原得清晰图像。实验结果表明,本文算法可以更好地复原退化图像,在主观视觉和客观质量评价方面都有所提高。     

压缩感知理论与非凸优化方法研究

针对某些信号带宽较宽导致难以直接采样的问题,压缩感知理论提供了一种可行的低速采样方法。信号在特定变换域中拥有稀疏表示,通过低速采样得到少量的投影值,已经包含了重构所需的重要信息。利用压缩感知理论从投影值中重构出稀疏向量,进而重建原信号。同时介绍一种基于非凸优化的压缩感知重构算法。相比L1范数的凸优化和无稀疏约束的L2范数,非凸优化的Lp范数拥有对稀疏性更强的约束。实验结果表明,使用压缩感知理论可以显著降低对信号的采样速率,而使用非凸优化算法可以取得更好的重构效果。     

基于非负稀疏表示的SAR图像目标识别方法

针对合成孔径雷达(SAR)图像目标识别中存在物体遮挡的情况,该文提出一种基于非负稀疏表示的分类方法。通过分析L0范数和L1范数最小化在求解非负稀疏表示问题上的区别,证明在一定条件下,L1范数最小化方法除了保持解的稀疏性还能得到与输入信号更加相似的原子集合,因此也更加适用于分类问题中。在运动和静止目标获取与识别(MSTAR)数据集上的识别实验结果表明,采用L1范数的非负稀疏表示分类方法能达到较好的识别性能,并且相对传统方法对存在遮挡情况下的识别问题更稳健。     

L0-norm sparse representation based on modified genetic algorithm for face recognition

The typical sparse representation for classification (SRC) exploits the training samples to represent the test samples, and classifies the test samples based on the representation results. SRC is essentially an L₀-norm minimization problem which can theoretically yield the sparsest representation and lead to the promising classification performance. We know that it is difficult to directly resolve L₀-norm minimization problem by applying usual optimization method. To effectively address this problem, we propose the L₀-norm based SRC by exploiting a modified genetic algorithm (GA), termed GASRC, in this paper. The basic idea of GASRC is that it modifies the traditional genetic algorithm and then uses the modified GA (MGA) to select a part of the training samples to represent a test sample. Compared with the conventional SRC based on L₁-norm optimization, GASRC can achieve better classification performance. Experiments on several popular real-world databases show the good classification effectiveness of our approach.     

基于稀疏信号重构的近场源定位

针对近场源定位问题,提出了一种基于稀疏信号重构的定位方法.该方法通过约束稀疏信号的L1-范数求解优化问题,实现信源的定位.该方法采用一种新的方法约束噪声项系数以求解优化问题,无需噪声的先验知识.为了减小计算量,将近场源二维定位问题转化为两次一维参数估计.通过计算机仿真验证了该方法的性能.     

Coarse-to-fine outlier correction with applications in structure from motion

This paper presents a robust method for outlier detection and correction in structure from motion. Aiming at handling outliers together with missing data, the Discrete Cosine Transform (DCT) based Column Space Fitting (CSF) algorithm is extended and improved. The use of the DCT basis allows for a coarse-to-fine optimization strategy that reconstructs 3D scene geometry and camera motion by increasing the number of DCT basis vectors. With a certain DCT basis, an interior point based L₁-norm solver is used to successively estimate 3D scene structure. In addition, the fidelity of the estimated camera motion matrix is first integrated into an extension of Huber M-estimator to find outliers and to robustly estimate the update magnitude for each outlier. This fidelity can be measured by the effects of camera motion matrix on re-projection errors. Because the Huber M-estimator is only applicable to vector, we extend it into the matrix form. With the increase in the number of DCT basis vectors, outliers are corrected in a coarse-to-fine manner. Experiments on both synthetic and real image sequences confirm the effectiveness of the proposed method.     

L/sub 2/-sensitivity minimization of one- and two-dimensional state-space digital filters subject to L/sub 2/-scaling constraints

This paper investigates the problem of minimizing an L/sub 2/-sensitivity measure subject to L/sub 2/-norm dynamic-range scaling constraints for one-dimensional (1-D) as well as a class of two-dimensional (2-D) state-space digital filters, where the 2-D digital filters are described by a transposed structure of the Fornasini-Marchesini second local state-space model. In each case, a novel iterative technique is developed to solve the constraint optimization problem directly. The proposed solution methods are largely based on the use of a Lagrange function and some matrix-theoretic techniques. Numerical examples are presented to demonstrate the effectiveness of the proposed techniques.     

模拟电路故障诊断L1估计及其神经网络解法

基于精确罚函数法,提出了新的求解L₁范致问题最优解的神经网络方法,它避免了Kennedy和Chua(1988)网络罚因子较大时性态变坏问题。对Bandler(1982)提出的模拟电路故障诊断L₁范数法进行了改进,将线性约束L₁问题转化为非线性约束L₁问题,并用新的神经网络方法求解,计算量小。模拟实验表明,所提神经网络方法和改进的模拟电路故障诊断L₁范数方法是可行的。