In this paper, k block sparse vectors are studied, and the block l(1) - l(2) model is adopted. It is proved theoretically that when the block sparsity satisfies some conditions, the k block sparse vector can be accura...
详细信息
In this paper, k block sparse vectors are studied, and the block l(1) - l(2) model is adopted. It is proved theoretically that when the block sparsity satisfies some conditions, the k block sparse vector can be accurately recovered by the noise free block l(1) - l(2) model, and it can also be stably recovered by the noisy block l(1) - l(2) model. In the algorithm, we use the convex difference algorithm, and prove that the aggregation points of the sequence generated by the algorithm converge to the stable point of the objective function. We prove that when the parameter lambda > 0 is less than a certain number lambda(k), the aggregation points of the sequence generated by the algorithm are block sparse. Finally, we conduct data experiments. The experiments show that when the vector is block sparse, the block l(1) - l(2) model can recover the unknown vector better than the traditional l(1) - l(2) model.
Convolutional neural networks (CNNs)-based deep features have been demonstrated with remarkable performance in various vision tasks, such as image classification and face verification. Compared with the hand-crafted d...
详细信息
Convolutional neural networks (CNNs)-based deep features have been demonstrated with remarkable performance in various vision tasks, such as image classification and face verification. Compared with the hand-crafted descriptors, deep features exhibit more powerful representation ability. Typically, higher layer features contain more semantic information, while lower layer features can provide more low-level description. In addition, it turns out that the fusion of different layer features will lead to superior performance. Here, we propose a novel approach for human ear identification by combining hierarchical deep features. First, hierarchical deep features are extracted from ear images using CNN pre-trained on large-scale data set. To enhance the feature representation and reduce the high dimension of deep features, the discriminant correlation analysis (dca) is adopted for fusing deep features from different layers for further improvement. Owing to the lack of ear images per person, the authors propose to transform the ear identification problem to the binary classification by composing pairwise samples and resolve it with the pairwise support vector machine (SVM). Experiments are conducted on four public databases: USTB I, USTB II, IIT Delhi I, and IIT Delhi II. The proposed method achieves promising recognition rate and exhibits decent performance compared with the state-of-the-art methods.
Numerous problems in signal processing and imaging, statistical learning and data mining, or computer vision can be formulated as optimization problems which consist in minimizing a sum of convex functions, not necess...
详细信息
Numerous problems in signal processing and imaging, statistical learning and data mining, or computer vision can be formulated as optimization problems which consist in minimizing a sum of convex functions, not necessarily differentiable, possibly composed with linear operators and that in turn can be transformed to split feasibility problems (SFP);see for example Censor and Elfving (Numer. algorithms 8, 221-239 1994). Each function is typically either a data fidelity term or a regularization term enforcing some properties on the solution;see for example Chaux et al. (SIAM J. Imag. Sci. 2, 730-762 2009) and references therein. In this paper, we are interested in split feasibility problems which can be seen as a general form of Q-Lasso introduced in Alghamdi et al. (2013) that extended the well-known Lasso of Tibshirani (J. R. Stat. Soc. Ser. B 58, 267-288 1996). Q is a closed convex subset of a Euclidean m-space, for some integer m >= 1, that can be interpreted as the set of errors within given tolerance level when linear measurements are taken to recover a signal/image via the Lasso. Inspired by recent works by Lou and Yan (2016), Xu (IEEE Trans. Neural Netw. Learn. Syst. 23, 1013-1027 2012), we are interested in a nonconvex regularization of SFP and propose three split algorithms for solving this general case. The first one is based on the DC (difference of convex) algorithm (dca) introduced by Pham Dinh Tao, the second one is nothing else than the celebrate forward-backward algorithm, and the third one uses a method introduced by Mine and Fukushima. It is worth mentioning that the SFP model a number of applied problems arising from signal/image processing and specially optimization problems for intensity-modulated radiation therapy (IMRT) treatment planning;see for example Censor et al. (Phys. Med. Biol. 51, 2353-2365, 2006).
Ear recognition problem is known as selecting whether two ear images belong to the same person or not, this consider as a challenge due to variation in lighting, background, pose, scale, and occlusion. This paper pres...
详细信息
ISBN:
(纸本)9781450353410
Ear recognition problem is known as selecting whether two ear images belong to the same person or not, this consider as a challenge due to variation in lighting, background, pose, scale, and occlusion. This paper presents an improvement method for unconstrained ear recognition problem based on local feature fusion, and further analyzes the performance and efficiency of discriminative local feature fusion for aligned and non-aligned ear images. Firstly, local discriminative features such as LPQ, HOG, LBP, POEM, BSIF and Gabor features are extracted from the ear images. Then, Discriminant Correlation Analysis (dca) is exploited for fusion and reduction dimension. Finally, support vector machine (SVM) is adopted for classification. Experiments are conducted on popular ear databases, USTB I, USTB II, and IIT Delhi II. Furthermore, we report an encouraging result on a difficult and challenging ear database called annotated web ear (AWE) that is collected from the wild. The experimental results show superior of proposed approach that can achieve a high performance for non-aligned images (AWE and USTB II datasets), on the other hand, unique local features can achieve promising recognition rates for aligned images, USTB I and IIT Delhi II datasets.
The focus of this paper is in Q-Lasso introduced in Alghamdi et al. (2013) which extended the Lasso by Tibshirani (1996). The closed convex subset Q belonging in a Euclidean m-space, for m is an element of IN, is the ...
详细信息
The focus of this paper is in Q-Lasso introduced in Alghamdi et al. (2013) which extended the Lasso by Tibshirani (1996). The closed convex subset Q belonging in a Euclidean m-space, for m is an element of IN, is the set of errors when linear measurements are taken to recover a signal/image via the Lasso. Based on a recent work by Wang (2013), we are interested in two new penalty methods for Q-Lasso relying on two types of difference of convex functions (DC for short) programming where the DC objective functions are the difference of l1 and l sigma q norms and the difference of l(1) and l(r) norms with r>1. By means of a generalized q-term shrinkage operator upon the special structure of l(sigma q) norm, we design a proximal gradient algorithm for handling the DC l(1)-l(sigma q) model. Then, based on the majorization scheme, we develop a majorized penalty algorithm for the DC l(1)-l(r) model. The convergence results of our new algorithms are presented as well. We would like to emphasize that extensive simulation results in the case Q={b} show that these two new algorithms offer improved signal recovery performance and require reduced computational effort relative to state-of-the-art l(1) and l(p) (p is an element of(0,1)) models, see Wang (2013). We also devise two DC algorithms on the spirit of a paper where exact DC representation of the cardinality constraint is investigated and which also used the largest-q norm of l(sigma q) and presented numerical results that show the efficiency of our DC algorithm in comparison with other methods using other penalty terms in the context of quadratic programing, see Jun-ya et al. (2017).
暂无评论