Structured Sparse Coding for Classification via Reweighted \(\ell _{2,1}\) Minimization
In recent years, sparse coding has been used in a wide range of applications including classification and recognition. Different from many other applications, the sparsity pattern of features in many classification tasks are structured and constrained in
- PDF / 697,238 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 76 Downloads / 235 Views
South China University of Technology, Guangzhou 510006, China [email protected], [email protected], [email protected] 2 National University of Singapore, Singapore 119076, Singapore [email protected] Abstract. In recent years, sparse coding has been used in a wide range of applications including classification and recognition. Different from many other applications, the sparsity pattern of features in many classification tasks are structured and constrained in some feasible domain. In this paper, we proposed a re-weighted 2,1 norm based structured sparse coding method to exploit such structures in the context of classification and recognition. In the proposed method, the dictionary is learned by imposing the class-specific structured sparsity on the sparse codes associated with each category, which can bring noticeable improvement on the discriminability of sparse codes. An alternating iterative algorithm is presented for the proposed sparse coding scheme. We evaluated our method by applying it to several image classification tasks. The experiments showed the improvement of the proposed structured sparse coding method over several existing discriminative sparse coding methods on tested data sets. Keywords: Sparse coding classification
1
·
Reweighted 2,1 minimization
·
Image
Introduction
In recent years, sparse model has been an important tool with a wide range of applications. Sparse modeling assumes that signals of interest can be succinctly expressed under some suitable system in a linear manner. The elements used for expressing signals are often referred as atoms and the collection of all such atoms is called a dictionary for sparse modeling. The computational method for sparse modeling is called sparse coding, which aims at finding a dictionary, Yong Xu would like to thank the supports by National Nature Science Foundations of China (61273255 and 61070091), Engineering and Technology Research Center of Guangdong Province for Big Data Analysis and Processing ([2013]1589-1-11), Project of High Level Talents in Higher Institution of Guangdong Province (20132050205-47) and Guangdong Technological Innovation Project (2013KJCX0010). Yuping Sun would like to thank the support by China Scholarship Council Program. Yuhui Quan and Yu Luo would like to thank the partial support by Singapore MOE Research Grant R-146-000-178-112. c Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part I, CCIS 546, pp. 189–199, 2015. DOI: 10.1007/978-3-662-48558-3 19
190
Y. Xu et al.
as well as the sparse coefficients, from input signals. This sparse scheme, which rigorously pursues the sparsity of the codes, works quite well in image processing and restoration. However, it is not enough to achieve high discriminability for classification and recognition tasks without exploiting extra structural information existing in signals. While some recent approaches have attempted to pursue structured sparsity for classification either explicitly or implicitly, the disadvantages of these approaches are obvious. For exam
Data Loading...
