Innovative PEPS Tensor Network Decomposition for Enhanced Higher Order Data Recovery

Abstract

Tensor decompositions (TDs) have demonstrated significant potential across various domains of science and engineering. Despite its thorough examination in quantum physics, the projected entangled pair state (PEPS) tensor network has not been extensively explored in the field of tensor completion (TC). In this study, we introduce an innovative PEPS tensor network decomposition algorithm that transforms an Nth-order tensor into a PEPS representation through

N-

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$N-$$\end{document}

1 sequential singular value decompositions (SVDs), considering the multilayered architecture of the PEPS tensor network. The accuracy of this algorithm can be finely tuned by adjusting the rank of the PEPS model. We apply the PEPS tensor network decomposition to address high-order TC challenges and develop an efficient proximal alternating minimization-based optimization algorithm to uncover latent factors within incomplete tensors, thereby filling missing entries. Comparative experiments on synthetic data and color image completion validate the efficacy of our proposed approaches, demonstrating significant improvements over existing methods based on CANDECOMP/PARAFAC (CP), Tucker, tensor train (TT), and tensor ring (TR) decompositions. The code and associated datasets are available at: https://github.com/rfhuang211/PEPS-TC.

Keywords

Projected entangled pair state (PEPS) / Tensor decomposition (TD) / Tensor completion (TC) / Singular value decomposition (SVD) / Proximal alternating minimization (PAM) / 65K05 / 90C46

References

[1] Acar E, Dunlavy DM, Kolda TG, Mørup M. Scalable tensor factorizations for incomplete data. Chemom. Intell. Lab. Syst., 2011, 106: 41-56

[2] Bengua, J.A., Phien, H.N., Tuan, H.D., Do, M.N.: Efficient tensor completion for color image and video recovery: low-rank tensor train. IEEE Trans. Image Process. 26, 2466–2479 (2017). https://doi.org/10.1109/tip.2017.2672439

[3] Bu Y, Zhao Y, Chan JC-W. Mixed norm regularized models for low-rank tensor completion. Inf. Sci., 2024, 670 120630

[4] Carroll, J.D., Chang, J.-J.: Analysis of individual differences in multidimensional scaling via an n\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$n$$\end{document}-way generalization of “Eckart-Young” decomposition. Psychometrika 35, 283–319 (1970). https://doi.org/10.1007/bf02310791

[5] Che H, Pan B, Leung M-F, Cao Y, Yan Z. Tensor factorization with sparse and graph regularization for fake news detection on social networks. IEEE Trans. Comput. Soc. Syst., 2023, 11: 1-11

[6] Chen C, Wu Z-B, Chen Z-T, Zheng Z-B, Zhang X-J. Auto-weighted robust low-rank tensor completion via tensor-train. Inf. Sci., 2021, 567: 100-115

[7] Eckart C, Young G. The approximation of one matrix by another of lower rank. Psychometrika, 1936, 1: 211-218

[8] Fernandes S, Fanaee-T H, Gama J. Tensor decomposition for analysing time-evolving social networks: an overview. Artif. Intell. Rev., 2020, 54: 2891-2916

[9] Huang, R., Liu, S., Zhang, X., Liu, Y., Zhao, Y.: Projected entangled pair state tensor network for colour image and video completion. In: Tanveer, M., Agarwal, S., Ozawa, S., Ekbal, A., Jatowt, A. (eds) Neural Information Processing. ICONIP 2022. Communications in Computer and Information Science, pp. 26–38 (2023). https://doi.org/10.1007/978-981-99-1645-0_3

[10] Karatzoglou, A., Amatriain, X., Baltrunas, L., Oliver, N.: Multiverse recommendation: N\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$N$$\end{document}-dimensional tensor factorization for context-aware collaborative filtering. In: Proceedings of the Fourth ACM Conference on Recommender Systems, pp. 79–86 (2010). https://doi.org/10.1145/1864708.1864727

[11] Kilmer ME, Braman K, Hao N, Hoover RC. Third-order tensors as operators on matrices: a theoretical and computational framework with applications in imaging. SIAM J. Matrix Anal. Appl., 2013, 34: 148-172

[12] Kolda TG, Bader BW. Tensor decompositions and applications. SIAM Rev., 2009, 51: 455-500

[13] Liu, D., Sacchi, M.D., Chen, W.: Efficient tensor completion methods for 5-D seismic data reconstruction: low-rank tensor train and tensor ring. IEEE Trans. Geosci. Remote Sens. 60, 1–17 (2022). https://doi.org/10.1109/tgrs.2022.3179275. (Accessed 2024-11-23)

[14] Liu, J., Musialski, P., Wonka, P., Ye, J.: Tensor completion for estimating missing values in visual data. IEEE Trans. Pattern Anal. Mach. Intell. 35, 208–220 (2013). https://doi.org/10.1109/tpami.2012.39

[15] Long Z, Liu Y, Chen L, Zhu C. Low rank tensor completion for multiway visual data. Signal Process., 2019, 155: 301-316

[16] Long, Z., Zhu, C., Liu, J., Liu, Y.: Bayesian low rank tensor ring for image recovery. IEEE Trans. Image Process. 30, 3568–3580 (2021). https://doi.org/10.1109/tip.2021.3062195

[17] Lyu C, Lu Q-L, Wu X, Antoniou C. Tucker factorization-based tensor completion for robust traffic data imputation. Transp. Res. Part C Emerg. Technol., 2024, 160 104502

[18] Orús, R.: A practical introduction to tensor networks: matrix product states and projected entangled pair states. Ann. Phys. 349, 117–158 (2014). https://doi.org/10.1016/j.aop.2014.06.013

[19] Oseledets IV. Tensor-train decomposition. SIAM J. Sci. Comput., 2011, 33: 2295-2317

[20] Perez-Garcia D, Verstraete F, Wolf MM, Cirac JI. Matrix product state representations. Quantum Inf. Comput., 2007, 7(5): 401-430

[21] Qiu Y, Zhou G, Zhao Q, Xie S. Noisy tensor completion via low-rank tensor ring. IEEE Trans. Neural Netw. Learn. Syst., 2022, 35: 1127-1141

[22] Said AB, Erradi A. Spatiotemporal tensor completion for improved urban traffic imputation. IEEE Trans. Intell. Transp. Syst., 2022, 23: 6836-6849

[23] Tucker LR. Some mathematical notes on three-mode factor analysis. Psychometrika, 1966, 31: 279-311

[24] Vidal G. Efficient classical simulation of slightly entangled quantum computations. Phys. Rev. Lett., 2003, 91 147902

[25] Wang Q, Han D. A generalized inertial proximal alternating linearized minimization method for nonconvex nonsmooth problems. Appl. Numer. Math., 2023, 189: 66-87

[26] Wang, W., Aggarwal, V., Aeron, S.: Tensor completion by alternating minimization under the tensor train (TT) model. arXiv:1609.05587 (2016)

[27] Wang, W., Aggarwal, V., Aeron, S.: Efficient low rank tensor ring completion. In: IEEE International Conference on Computer Vision, pp. 5698–5706 (2017). https://doi.org/10.1109/ICCV.2017.607

[28] Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P.: Image quality assessment: from error visibility to structural similarity. IEEE Trans. Image Process. 13, 600–612 (2004). https://doi.org/10.1109/tip.2003.819861

[29] Wu, Z.-C., Huang, T.-Z., Deng, L.-J., Dou, H.-X., Meng, D.: Tensor wheel decomposition and its tensor completion application. In: Proceedings of the 36th International Conference on Neural Information Processing Systems, pp. 27008–27020 (2024)

[30] Xu Y, Fu L, Niu X, Chen X, Zhang M. Three-dimensional seismic data reconstruction based on fully connected tensor network decomposition. IEEE Trans. Geosci. Remote Sens., 2023, 61: 1-11

[31] Yuan, L., Cao, J., Zhao, X., Wu, Q., Zhao, Q.: Higher-dimension tensor completion via low-rank tensor ring decomposition. In: 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), pp. 1071–1076 (2018). https://doi.org/10.23919/APSIPA.2018.8659708

[32] Yuan, L., Li, C., Mandic, D., Cao, J., Zhao, Q.: Tensor ring decomposition with rank minimization on latent space: an efficient approach for tensor completion. Proceedings of the AAAI Conference on Artificial Intelligence 33(01), 9151–9158 (2019). https://doi.org/10.1609/aaai.v33i01.33019151

[33] Yuan, L., Zhao, Q., Cao, J.: Completion of high order tensor data with missing entries via tensor-train decomposition. In: Liu, D., Xie, S., Li, Y., Zhao, D., El-Alfy, ES. (eds) Neural Information Processing. ICONIP 2017. Lecture Notes in Computer Science, pp. 222–229. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-70087-8_24

[34] Zhang, X., Gong, Y., Qiao, C., Jing, W.: Multiview deep learning based on tensor decomposition and its application in fault detection of overhead contact systems. Vis. Comput. 38, 1457–1467 (2022). https://doi.org/10.1007/s00371-021-02080-y

[35] Zhang, Z., Ely, G., Aeron, S., Hao, N., Kilmer, M.: Novel methods for multilinear data completion and de-noising based on tensor-SVD. In: 2014 IEEE Conference on Computer Vision and Pattern Recognition, pp. 3842–3849 (2014). https://doi.org/10.1109/CVPR.2014.485

[36] Zhao, Q., Zhang, L., Cichocki, A.: Bayesian CP factorization of incomplete tensors with automatic rank determination. IEEE Trans. Pattern Anal. Mach. Intell. 37, 1751–1763 (2015). https://doi.org/10.1109/tpami.2015.2392756

[37] Zhao, Q., Zhou, G., Xie, S., Zhang, L., Cichocki, A.: Tensor ring decomposition. arXiv:1606.05535 (2016)

[38] Zheng, Y.-B., Huang, T., Zhao, X., Zhao, Q., Jiang, T.-X.: Fully-connected tensor network decomposition and its application to higher-order tensor completion. Proceedings of the AAAI Conference on Artificial Intelligence 35(12), 11071–11078 (2021). https://doi.org/10.1609/aaai.v35i12.17321

Funding

the Strategic Priority Research Program of the Chinese Academy of Sciences(XDB0500101)

RIGHTS & PERMISSIONS

Shanghai University

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.