Discriminative Local Sparse Representation by Robust Adaptive Dictionary Pair Learning

IEEE transactions on neural networks and learning systems, pp. 1-15, 2020.

Cited by: 7|Bibtex|Views140|Links
WOS
Keywords:
K-singular value decompositionlocal sparseImage recognitiondiscriminative K-SVDstructured DLMore(21+)
Weibo:
We propose a structured robust adaptive dictionary pair learning framework for the discriminative sparse representation learning

Abstract:

In this article, we propose a structured robust adaptive dictionary pair learning (RA-DPL) framework for the discriminative sparse representation (SR) learning. To achieve powerful representation ability of the available samples, the setting of RA-DPL seamlessly integrates the robust projective DPL, locality-adaptive SRs, and discriminati...More

Code:

Data:

0
Introduction
  • E FFECTIVE image representation and classification via dictionary learning (DL) have received much attention in recent years and have been successfully applied to a variety of real-world emerging applications in a wide range of areas, such as computer vision [13], image denoising and compression [14], visual image classification [1]–[12], [38]–[41], and so on.
  • Real application data are usually corrupted by various noise and errors [2]; second, this kind of dictionary usually has a large size to make the process of coefficients coding inefficient [1], [2]
  • To address these issues, lots of efforts have devoted to the research of learning the compact and over-complete dictionaries in the area of representation learning [1]–[12], [47]–[54].
  • The structured DL (SDL) mechanism will be further investigated
Highlights
  • E FFECTIVE image representation and classification via dictionary learning (DL) have received much attention in recent years and have been successfully applied to a variety of real-world emerging applications in a wide range of areas, such as computer vision [13], image denoising and compression [14], visual image classification [1]–[12], [38]–[41], and so on
  • The averaged results are described in Table VI, from which we can see that: 1) the recognition accuracy increases as the training number increases and 2) our robust adaptive dictionary pair learning is superior to its competitors in most cases, and the main reason for the improvement by robust adaptive dictionary pair learning can be attributed to keeping the local neighborhood information and its robust adaptive dictionary pair learning scheme
  • More supervision information is beneficial to higher accuracies as the number of training data is increased; 2) robust adaptive dictionary pair learning is superior to its competitors
  • We find that: 1) dictionary pair learning, analysis discriminative DL, and our robust adaptive dictionary pair learning are more efficient than label-consistent K-SVD, discriminative K-SVD, and low-rank shared DL in general
  • We have proposed a robust adaptive projective dictionary pair learning framework for the discriminative local sparse data representations
  • Our model improves the representation and discriminating abilities of the existing projective dictionary pair learning from several aspects, i.e., enhancing the robust properties of the learning system to noise and corruptions in data, encouraging the coding coefficients to hold the sparse properties by efficient embedding, integrating the structured reconstruction weighting to preserve the local neighborhood within the coefficients of each class in an adaptive way, and including a discriminating function to ensure the intraclass compactness and interclass separation over the coefficients at the same time
Results
  • EXPERIMENTAL RESULTS AND ANALYSIS

    The authors mainly evaluate the RA-DPL for the data representation and classification.
  • 4) Face Recognition on UMIST Database: This database contains 1012 images of 20 individuals, and each individual is shown in a range of pose from profile to fontal views [27]
  • In this simulation, the authors randomly select five images per class for training and use other images for testing.
  • Results on AR Face Database
  • In this simulation, the authors randomly choose five and ten images per person for training and use the rest for testing.
  • LRSDL, DPL, and ADDL can work well by delivering better results than other remaining methods in most cases
Conclusion
  • The authors have proposed a robust adaptive projective DPL framework for the discriminative local sparse data representations.
  • The authors' model improves the representation and discriminating abilities of the existing projective DPL from several aspects, i.e., enhancing the robust properties of the learning system to noise and corruptions in data, encouraging the coding coefficients to hold the sparse properties by efficient embedding, integrating the structured reconstruction weighting to preserve the local neighborhood within the coefficients of each class in an adaptive way, and including a discriminating function to ensure the intraclass compactness and interclass separation over the coefficients at the same time.
  • Extending the method to the deep DL scenario [55], [56] and evaluating it on large-scale data sets will be discussed
Summary
  • Introduction:

    E FFECTIVE image representation and classification via dictionary learning (DL) have received much attention in recent years and have been successfully applied to a variety of real-world emerging applications in a wide range of areas, such as computer vision [13], image denoising and compression [14], visual image classification [1]–[12], [38]–[41], and so on.
  • Real application data are usually corrupted by various noise and errors [2]; second, this kind of dictionary usually has a large size to make the process of coefficients coding inefficient [1], [2]
  • To address these issues, lots of efforts have devoted to the research of learning the compact and over-complete dictionaries in the area of representation learning [1]–[12], [47]–[54].
  • The structured DL (SDL) mechanism will be further investigated
  • Results:

    EXPERIMENTAL RESULTS AND ANALYSIS

    The authors mainly evaluate the RA-DPL for the data representation and classification.
  • 4) Face Recognition on UMIST Database: This database contains 1012 images of 20 individuals, and each individual is shown in a range of pose from profile to fontal views [27]
  • In this simulation, the authors randomly select five images per class for training and use other images for testing.
  • Results on AR Face Database
  • In this simulation, the authors randomly choose five and ten images per person for training and use the rest for testing.
  • LRSDL, DPL, and ADDL can work well by delivering better results than other remaining methods in most cases
  • Conclusion:

    The authors have proposed a robust adaptive projective DPL framework for the discriminative local sparse data representations.
  • The authors' model improves the representation and discriminating abilities of the existing projective DPL from several aspects, i.e., enhancing the robust properties of the learning system to noise and corruptions in data, encouraging the coding coefficients to hold the sparse properties by efficient embedding, integrating the structured reconstruction weighting to preserve the local neighborhood within the coefficients of each class in an adaptive way, and including a discriminating function to ensure the intraclass compactness and interclass separation over the coefficients at the same time.
  • Extending the method to the deep DL scenario [55], [56] and evaluating it on large-scale data sets will be discussed
Tables
  • Table1: ROBUST ADAPTIVE PROJECTIVE DPL
  • Table2: RECOGNITION RESULTS USING RANDOM FACE FEATURES ON AR
  • Table3: RECOGNITION RESULTS USING CONVOLUTION FEATURES ON AR
  • Table4: DESCRIPTIONS OF USED REAL-WORLD IMAGE DATABASES
  • Table5: RECOGNITION RESULTS USING RANDOM FACE FEATURES ON YALEB
  • Table6: RECOGNITION RESULTS ON THE ETH80 OBJECT DATABASE
  • Table7: RECOGNITION RESULTS USING RANDOM FACE FEATURES ON CMU PIE
  • Table8: RECOGNITION RESULTS USING CONVOLUTION FEATURES ON CMU PIE
  • Table9: COMPARISON OF RECOGNITION RESULTS ON CMU PIE, UMIST, FIFTEEN SCENES, AND ETH80 UNDER DIFFERENT PARAMETERS
  • Table10: RECOGNITION RESULTS ON THE UMIST DATABASE
Download tables as Excel
Related work
  • In this section, we briefly review the related methods that are closely related to our formulation.

    A. Overall DL (ODL)

    Let X = [x1, · · · xl, · · · xN ] ∈ Rn×N be a set of training samples from c classes, where n is the original dimensionality and N is the number of samples. Then, ODL learns a reconstructive dictionary D of K atoms to deliver the SR S over the data X by the following general problem: D, S = arg min D,S X − D S 2F + λSp (1)

    where · 2F denotes the reconstruction error over
Funding
  • This work was supported in part by the National Natural Science Foundation of China under Grant 61672365, Grant 61732008, Grant 61725203, Grant 61622305, Grant 61871444, and Grant 61806035, and in part by the Fundamental Research Funds for the Central Universities of China under Grant JZ2019HGPA0102. (Corresponding author: Zhao Zhang.)
Reference
  • Z. Jiang, Z. Lin, and L. S. Davis, “Label consistent K-SVD: Learning a discriminative dictionary for recognition,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 35, no. 11, pp. 2651–2664, Nov. 2013.
    Google ScholarLocate open access versionFindings
  • Z. Zhang, F. Li, T. W. S. Chow, L. Zhang, and S. Yan, “Sparse codes auto-extractor for classification: A joint embedding and dictionary learning framework for representation,” IEEE Trans. Signal Process., vol. 64, no. 14, pp. 3790–3805, Jul. 2016.
    Google ScholarLocate open access versionFindings
  • M. Aharon, M. Elad, and A. Bruckstein, “K-SVD: An algorithm for designing overcomplete dictionaries for sparse representation,” IEEE Trans. Signal Process., vol. 54, no. 11, pp. 4311–4322, Nov. 2006.
    Google ScholarLocate open access versionFindings
  • J. Wright, A. Y. Yang, A. Ganesh, S. S. Sastry, and Y. Ma, “Robust face recognition via sparse representation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 31, no. 2, pp. 210–227, Feb. 2009.
    Google ScholarLocate open access versionFindings
  • Z. Li, Z. Lai, Y. Xu, J. Yang, and D. Zhang, “A locality-constrained and label embedding dictionary learning algorithm for image classification,” IEEE Trans. Neural Netw. Learn. Syst., vol. 28, no. 2, pp. 278–293, Feb. 2015.
    Google ScholarLocate open access versionFindings
  • Q. Zhang and B. Li, “Discriminative K-SVD for dictionary learning in face recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., San Francisco, CA, USA, Jun. 2010, pp. 2691–2698.
    Google ScholarLocate open access versionFindings
  • M. Yang, L. Zhang, X. Feng, and D. Zhang, “Sparse representation based Fisher discrimination dictionary learning for image classification,” Int. J. Comput. Vis., vol. 109, no. 3, pp. 209–232, Sep. 2014.
    Google ScholarLocate open access versionFindings
  • I. Ramirez, P. Sprechmann, and G. Sapiro, “Classification and clustering via dictionary learning with structured incoherence and shared features,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., San Francisco, CA, USA, Jun. 2010, pp. 3501–3508.
    Google ScholarLocate open access versionFindings
  • S. Gu, L. Zhang, W. Zuo, and X. Feng, “Projective dictionary pair learning for pattern classification,” in Proc. Conf. Neural Inf. Process. Syst., Montreal, QC, Canada, 2014, pp. 1–9.
    Google ScholarLocate open access versionFindings
  • Z. Zhang, W. Jiang, F. Li, M. Zhao, B. Li, and L. Zhang, “Structured latent label consistent dictionary learning for salient machine faults representation-based robust classification,” IEEE Trans. Ind. Informat., vol. 13, no. 2, pp. 644–656, Apr. 2017.
    Google ScholarLocate open access versionFindings
  • Z. Zhang et al., “Jointly learning structured analysis discriminative dictionary and analysis multiclass classifier,” IEEE Trans. Neural Netw. Learn. Syst., vol. 29, no. 8, pp. 3798–3814, Aug. 2018.
    Google ScholarLocate open access versionFindings
  • W. Jiang, Z. Zhang, F. Li, L. Zhang, M. Zhao, and X. Jin, “Joint label consistent dictionary learning and adaptive label prediction for semisupervised machine fault classification,” IEEE Trans. Ind. Informat., vol. 12, no. 1, pp. 248–256, Feb. 2016.
    Google ScholarLocate open access versionFindings
  • Q. Qiu, Z. L. Jiang, and R. Chellappa, “Sparse dictionary-based representation and recognition of action attributes,” in Proc. Int. Conf. Comput. Vis., Nov. 2011, pp. 707–714.
    Google ScholarLocate open access versionFindings
  • M. Elad, “Image denoising via sparse and redundant representations over learned dictionaries,” IEEE Trans. Image Process., vol. 54, no. 12, pp. 3736–3745, Dec. 2006.
    Google ScholarLocate open access versionFindings
  • R. Rubinstein, A. M. Bruckstein, and M. Elad, “Dictionaries for sparse representation modeling,” Proc. IEEE, vol. 98, no. 6, pp. 1045–1057, Jun. 2010.
    Google ScholarLocate open access versionFindings
  • L. Qiao, S. Chen, and X. Tan, “Sparsity preserving projections with applications to face recognition,” Pattern Recognit., vol. 43, no. 1, pp. 331–341, 2010.
    Google ScholarLocate open access versionFindings
  • Z. Zhang, Y. Zhang, F. Li, M. Zhao, L. Zhang, and S. Yan, “Discriminative sparse flexible manifold embedding with novel graph for robust visual representation and label propagation,” Pattern Recognit., vol. 61, pp. 492–510, Jan. 2017.
    Google ScholarLocate open access versionFindings
  • D. Cai, X. He, J. Han, and T. S. Huang, “Graph regularized nonnegative matrix factorization for data representation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 33, no. 8, pp. 1548–1560, Aug. 2011.
    Google ScholarLocate open access versionFindings
  • Z. Jiang, G. Zhang, and L. S. Davis, “Submodular dictionary learning for sparse coding,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Jun. 2012, pp. 3418–3425.
    Google ScholarLocate open access versionFindings
  • J. de Leeuw, “Block-relaxation algorithms in statistics,” in Information Systems and Data Analysis. Berlin, Germany: Springer, 1994, pp. 308– 324.
    Google ScholarFindings
  • R. E. Wendell and A. P. Hurter, Jr., “Minimization of a non-separable objective function subject to disjoint constraints,” Oper. Res., vol. 24, no. 4, pp. 643–657, 1976.
    Google ScholarLocate open access versionFindings
  • J. Gorski, F. Pfeuffer, and K. Klamroth, “Biconvex sets and optimization with biconvex functions: A survey and extensions,” Math. Methods Oper. Res., vol. 66, no. 3, pp. 373–407, Dec. 2007.
    Google ScholarLocate open access versionFindings
  • M. Yang, D. Dai, L. Shen, and L. Van Gool, “Latent dictionary learning for sparse representation based classification,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Jun. 2014, pp. 4138–4145.
    Google ScholarLocate open access versionFindings
  • T. H. Vu and V. Monga, “Fast low-rank shared dictionary learning for image classification,” IEEE Trans. Image Process., vol. 26, no. 11, pp. 5160–5175, Nov. 2017.
    Google ScholarLocate open access versionFindings
  • F. S. Samaria and A. C. Harter, “Parameterisation of a stochastic model for human face identification,” in Proc. IEEE Workshop Appl. Comput. Vis., Dec. 1994, pp. 138–142.
    Google ScholarLocate open access versionFindings
  • A. S. Georghiades, P. N. Belhumeur, and D. J. Kriegman, “From few to many: Illumination cone models for face recognition under variable lighting and pose,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 23, no. 6, pp. 643–660, Jun. 2001.
    Google ScholarLocate open access versionFindings
  • D. B. Graham and N. M. Allinson, “Characterising virtual eigensignatures for general purpose face recognition,” in Face Recognition: From Theory to Applications (NATO ASI Series), vol.
    Google ScholarLocate open access versionFindings
  • 163. Berlin, Germany: Springer, 1998, pp. 446–456.
    Google ScholarFindings
  • [28] A. Martinez, “The AR face database,” Comput. Vis. Center, Auton. Univ. Barcelona, Bellaterra, Spain, CVC Tech. Rep. 24, 1998.
    Google ScholarFindings
  • [29] T. Sim, S. Baker, and M. Bsat, “The CMU pose, illumination, and expression (PIE) database,” in Proc. 5th IEEE Int. Conf. Autom. Face Gesture Recognit., May 2002, pp. 53–58.
    Google ScholarLocate open access versionFindings
  • [30] B. Leibe and B. Schiele, “Analyzing appearance and contour based methods for object categorization,” in Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., Madison, WI, USA, Jun. 2003, p. II-409.
    Google ScholarLocate open access versionFindings
  • [31] S. Lazebnik, C. Schmid, and J. Ponce, “Beyond bags of features: Spatial pyramid matching for recognizing natural scene categories,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., New York, NY, USA, Jun. 2006, pp. 2169–2178.
    Google ScholarLocate open access versionFindings
  • [32] S. Kong and D. Wang, “A dictionary learning approach for classification: Separating the particularity and the commonality,” in Proc. Eur. Conf. Comput. Vis. Berlin, Germany: Springer, 2012, pp. 186–199.
    Google ScholarLocate open access versionFindings
  • [33] K. Fukunaga, Introduction to Statistical Pattern Recognition, 2nd ed. Boston, MA, USA: Academic, 2002.
    Google ScholarFindings
  • [34] Z. Zhang, L. Jia, M. Zhao, G. Liu, M. Wang, and S. Yan, “Kernelinduced label propagation by mapping for semi-supervised classification,” IEEE Trans. Big Data, vol. 5, no. 2, pp. 148–165, Jun. 2018.
    Google ScholarLocate open access versionFindings
  • [35] D. D. Lee and H. S. Seung, “Algorithms for non-negative matrix factorization,” in Proc. Adv. Neural Inf. Process. Syst., vol.
    Google ScholarLocate open access versionFindings
  • 13. Cambridge, MA, USA: MIT Press, 2001, pp. 556–562.
    Google ScholarFindings
  • [36] B. Chen, J. Li, B. Ma, and W. Gang, “Discriminative dictionary pair learning based on differentiable support vector function for visual recognition,” Neurocomputing, vol. 272, pp. 306–313, Jan. 2018.
    Google ScholarLocate open access versionFindings
  • [37] B. Weyrauch, J. Huang, B. Heisele, and V. Blanz, “Component-based face recognition with 3D morphable models,” in Proc. Conf. Comput. Vis. Pattern Recognit. Workshop, Washington, DC, USA, 2004, p. 85.
    Google ScholarLocate open access versionFindings
  • [38] V. Singhal, H. K. Aggarwal, S. Tariyal, and A. Majumdar, “Discriminative robust deep dictionary learning for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens., vol. 55, no. 9, pp. 5274–5283, Sep. 2017.
    Google ScholarLocate open access versionFindings
  • [39] W. Fu, S. Li, L. Fang, and J. A. Benediktsson, “Contextual online dictionary learning for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens., vol. 56, no. 3, pp. 1336–1347, Mar. 2017.
    Google ScholarLocate open access versionFindings
  • [40] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” in Proc. Adv. Neural Inf. Process. Syst., 2012, pp. 1106–1114.
    Google ScholarLocate open access versionFindings
  • [41] J. Hu and Y.-P. Tan, “Nonlinear dictionary learning with application to image classification,” Pattern Recognit., vol. 75, pp. 282–291, Mar. 2018.
    Google ScholarLocate open access versionFindings
  • [42] C.-C. Hsu and C.-W. Lin, “CNN-based joint clustering and representation learning with feature drift compensation for large-scale image data,” IEEE Trans. Multimedia, vol. 20, no. 2, pp. 421–429, Feb. 2018.
    Google ScholarLocate open access versionFindings
  • [43] Y. Yang, H. T. Shen, Z. G. Ma, Z. Huang, and X. Zhou, “L2, 1-norm regularized discriminative feature selection for unsupervised learning,” in Proc. Int. Joint Conf. Artif. Intell., 2011, pp. 1–6.
    Google ScholarLocate open access versionFindings
  • [44] C. Hou, F. Nie, X. Li, D. Yi, and Y. Wu, “Joint embedding learning and sparse regression: A framework for unsupervised feature selection,” IEEE Trans. Cybern., vol. 44, no. 6, pp. 793–804, Jun. 2014.
    Google ScholarLocate open access versionFindings
  • [45] Y. Sun, Z. Zhang, W. Jiang, G. Liu, M. Wang, and S. Yan, “Robust discriminative projective dictionary pair learning by adaptive representations,” in Proc. 24th Int. Conf. Pattern Recognit., Beijing, China, 2018, pp. 621–626.
    Google ScholarLocate open access versionFindings
  • [46] Z. Zhang, Y. Zhang, G. Liu, J. Tang, S. Yan, and M. Wang, “Joint label prediction based semi-supervised adaptive concept factorization for robust data representation,” IEEE Trans. Knowl. Data Eng., to be published.
    Google ScholarLocate open access versionFindings
  • [47] G. Liu, Z. Zhang, Q. Liu, and H. Xiong, “Robust subspace clustering with compressed data,” IEEE Trans. Image Process., vol. 28, no. 10, pp. 5161–5170, Oct. 2019.
    Google ScholarLocate open access versionFindings
  • [48] L. Wang et al., “Robust auto-weighted projective low-rank and sparse recovery for visual representation,” Neural Netw., vol. 117, pp. 201–215, May 2019.
    Google ScholarLocate open access versionFindings
  • [49] S. Li, M. Shao, and Y. Fu, “Person re-identification by cross-view multi-level dictionary learning,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 40, no. 12, pp. 2963–2977, Dec. 2017.
    Google ScholarLocate open access versionFindings
  • [50] J. J. Thiagarajan, K. N. Ramamurthy, and A. Spanias, “Learning stable multilevel dictionaries for sparse representations,” IEEE Trans. Neural Netw. Learn. Syst., vol. 26, no. 9, pp. 1913–1926, Sep. 2015.
    Google ScholarLocate open access versionFindings
  • [51] K. Li, Z. Ding, S. Li, and Y. Fu, “Toward resolution-invariant person reidentification via projective dictionary learning,” IEEE Trans. Neural Netw. Learn. Syst., vol. 30, no. 6, pp. 1896–1907, Jun. 2019, doi: 10. 1109/TNNLS.2018.2875429.
    Locate open access versionFindings
  • [52] Z. Zhang et al., “Joint subspace recovery and enhanced locality driven robust flexible discriminative dictionary learning,” IEEE Trans. Circuits Syst. Video Technol., to be published, doi: 10.1109/TCSVT.2019.2923007.
    Locate open access versionFindings
  • [53] Z. Zhang, W. Jiang, Z. Zhang, S. Li, G. Liu, and J. Qin, “Scalable blockdiagonal locality-constrained projective dictionary learning,” in Proc. Int. Joint Conf. Artif. Intell., Macao, China, Aug. 2019, pp. 4376–4382.
    Google ScholarLocate open access versionFindings
  • [54] Z. Zhang, Y. Sun, Z. Zhang, Y. Wang, G. Liu, and M. Wang, “Learning structured twin-incoherent twin-projective latent dictionary pairs for classification,” in Proc. 19th IEEE Int. Conf. Data Mining, Beijing, China, Oct. 2019, pp. 1–10.
    Google ScholarLocate open access versionFindings
  • [55] Z. Ding, M. Shao, and Y. Fu, “Deep robust encoder through locality preserving low-rank dictionary,” in Proc. Eur. Conf. Comput. Vis., 2016, pp. 567–582.
    Google ScholarLocate open access versionFindings
  • [56] S. Wang, Z. Ding, and Y. Fu, “Marginalized denoising dictionary learning with locality constraint,” IEEE Trans. Image Process., vol. 27, no. 1, pp. 500–510, Jan. 2018.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments