Tensor Completion Algorithms in Big Data Analytics
Open accessThe DOI matches its Crossref record; 154 of the 154 references it cites were found in a registry or an index.
Abstract
Tensor completion is a problem of filling the missing or unobserved entries of partially observed tensors. Due to the multidimensional character of tensors in describing complex datasets, tensor completion algorithms and their applications have received wide attention and achievement in areas like data mining, computer vision, signal processing, and neuroscience. In this survey, we provide a modern overview of recent advances in tensor completion algorithms from the perspective of big data analytics characterized by diverse variety, large volume, and high velocity. We characterize these advances from the following four perspectives: general tensor completion algorithms, tensor completion with auxiliary information (variety), scalable tensor completion algorithms (volume), and dynamic tensor completion algorithms (velocity). Further, we identify several tensor completion applications on real-world data-driven problems and present some common experimental frameworks popularized in the literature along with several available software repositories. Our goal is to summarize these popular methods and introduce them to researchers and practitioners for promoting future research and applications. We conclude with a discussion of key challenges and promising research directions in this community for future exploration.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Donoho, D. L. (2006) · IEEE Transactions on Information Theory · doi:10.1109/tit.2006.871582 · Cited by 23,994
Candes, E. J., Romberg, J., & Tao, T. (2006) · IEEE Transactions on Information Theory · doi:10.1109/tit.2005.862083 · Cited by 15,639 · cited 1×
“mes and show the rapid adaptivity of their method OLSTEC to the changed background. Other video completion examples could be found in [187,220]. 7.5 Compressed Sensing Compressed sensing initiated by [27, 43] in 2006 allows the recovery of a sparse signal from a small number of nonadaptive, linear random measurements…”
MethodInfluentialMatrix Factorization Techniques for Recommender Systems
Koren, Y., Bell, R., & Volinsky, C. (2009) · Computer · doi:10.1109/mc.2009.263 · Cited by 11,673 · cited 1×
“he tensor completion problem in recent years, to the best of our knowledge, we still lack a comprehensive survey to capture the many advances across domains. Existing surveys rest on the matrix level [88, 106, 115, 181] or focus on related topics, e.g., tensor decomposition [4 ,36 51 67 74 99 103 109 110 130 148 167 1…”
BackgroundTensor Decompositions and Applications
Kolda, T. G., & Bader, B. W. (2009) · SIAM Review · doi:10.1137/07070111x · Cited by 10,301 · cited 5×
“e of big data analytics. 3.1 Decomposition Based Approaches Tensor decompositions or factorizations are powerful tools for extracting meaningful, latent structures in heterogeneous, multi-aspect data [84, 121]. Since real-world datasets are o›en raw and incomplete, many early researches are conducted on tensor decompo…”
BackgroundMethodInfluentialSignal Recovery From Random Measurements Via Orthogonal Matching Pursuit
Tropp, J. A., & Gilbert, A. C. (2007) · IEEE Transactions on Information Theory · doi:10.1109/tit.2007.909108 · Cited by 9,653 · cited 1×
“s G i 2R I 1 ::: N via »G„X”…„i”=< G i;X >, where each tensor G i has i.i.d. standard normal entries. It enjoys sharp theoretical tools [138] and has been well-studied in signal processing •eld [195] especially in the compressed sensing area [43]. Another widely studied sampling measurement is random Fourier mea…”
BackgroundRobust principal component analysis?
Candès, E. J., Li, X., Ma, Y., & Wright, J. (2011) · Journal of the ACM · doi:10.1145/1970392.1970395 · Cited by 6,921 · cited 1×
“.3.2 Robust Tensor Completion. Robust data analysis plays an instrumental role in dealing with outliers, gross corruptions in completion problem [57]. Drawing upon the advances in robust PCA analysis [22], robust tensor completion is to complete a tensor B by separating it into a low-rank part X plus a sparse part E t…”
BackgroundInfluentialA Singular Value Thresholding Algorithm for Matrix Completion
Cai, J.-F., Candès, E. J., & Shen, Z. (2010) · SIAM Journal on Optimization · doi:10.1137/080738970 · Cited by 5,809 · cited 2×
“me of ADMM, fZ ng n and fY ng n could be iteratively updated as follows: Z n SVT αn η „A n Y n η ”; Y n Y n +η„Z n A n”; n = 1;2;:::;N; (24) where SVT αn η is the singular value thresholding operator [20] which is a shrinkage operator de•ned as SVT δ„A”= U„diagfσ i δg” +V>, where U„diagfσ ig 1ir”V>represents the…”
BackgroundCarroll, J. D., & Chang, J.-J. (1970) · Psychometrika · doi:10.1007/bf02310791 · Cited by 4,678
A Multilinear Singular Value Decomposition
De Lathauwer, L., De Moor, B., & Vandewalle, J. (2000) · SIAM Journal on Matrix Analysis and Applications · doi:10.1137/s0895479896305696 · Cited by 4,106 · cited 2×
“to handle missing values and conduct imputation. 3.1.2 Tucker Based Methods. Tucker decomposition is proposed by Tucker [158] and further developed by Kroonenberg et al. [92] and De Lathauwer et al. [35]. Given an Nth-order tensor X, its Tucker decomposition is de•ned as an approximation of a core tensor C 2RΠNn =1 R…”
BackgroundMethodInfluentialFactorization meets the neighborhood: a multifaceted collaborative filtering model
Koren, Y. (2008) · Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/1401890.1401944 · Cited by 3,852
A Survey of Collaborative Filtering Techniques
Su, X., & Khoshgoftaar, T. M. (2009) · Advances in Artificial Intelligence · doi:10.1155/2009/421425 · Cited by 3,602 · cited 1×
“Existing surveys rest on the matrix level [90, 109, 118, 188] or focus on related topics, e.”
BackgroundSome Mathematical Notes on Three-Mode Factor Analysis
Tucker, L. R. (1966) · Psychometrika · doi:10.1007/bf02289464 · Cited by 3,492 · cited 2×
“th EM-like approach when the missing ratio is large but could be hard to be optimized with existing projection-based algorithms. 3.1.2 Tucker-Based Methods. Tucker decomposition is proposed by Tucker [196] and further developed by Kroonenberg et al. [111] and De Lathauwer et al. [40]. Given an Nth-order tensor X, its…”
MethodInfluentialGuaranteed Minimum-Rank Solutions of Linear Matrix Equations via Nuclear Norm Minimization
Recht, B., Fazel, M., & Parrilo, P. A. (2010) · SIAM Review · doi:10.1137/070697835 · Cited by 3,471 · cited 1×
“For matrices with bounded operator norm, it has been shown to be the tightest lower bound of matrix rank function among all possible convex approximation [165].”
BackgroundRendle, S. (2010) · 2010 IEEE International Conference on Data Mining · doi:10.1109/icdm.2010.127 · Cited by 3,189 · cited 1×
“y incorporating interactions among di‡erent types of modes. (2) „ey have advantages in dealing with sparsity or cold start problem which are usually encountered in recommender systems as described in [159]. 7.1.3 Urban Computing. Urban computing deals with the human behaviors and mobilities with the help of computing…”
BackgroundInfluentialPARAFAC. Tutorial and applications
Bro, R. (1997) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/s0169-7439(97)00032-4 · Cited by 2,814 · cited 4×
“A famous dataset for tensor factorization and completion is the semi-realistic amino acid fluorescence data contributed by Bro and Andersson [14].”
MethodInfluentialOseledets, I. V. (2011) · SIAM Journal on Scientific Computing · doi:10.1137/090752286 · Cited by 2,636 · cited 1×
“The drawbacks of utilizing tensor CP rank and tensor n-rank encourage researchers to propose other types of ranks such as tensor-train (TT) rank [68, 151, 159] and tensor tubal rank [233].”
BackgroundGraph Regularized Nonnegative Matrix Factorization for Data Representation
Cai, D., He, X., Han, J., & Huang, T. S. (2011) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2010.231 · Cited by 2,451 · cited 1×
“tively non-negative threshold. For non-negative tensors with integer entries, Takeuchi and Ueda [150] use generalized Kullback-Leibler (gKL) divergence by analogy with several matrix-based approaches [16,37,165]. 3.3.2 Robust Tensor Completion. Robust data analysis plays an instrumental role in dealing with outliers,…”
MethodInfluentialIntroduction to Smooth Manifolds
Lee, J. M. (2012) · Graduate Texts in Mathematics · doi:10.1007/978-1-4419-9982-5 · Cited by 2,296
Kuybeda, O., Frank, G. A., Bartesaghi, A., Borgnia, M., Subramaniam, S., & Sapiro, G. (2013) · Journal of Structural Biology · doi:10.1016/j.jsb.2012.10.010 · Cited by 2,162
Tensor Completion for Estimating Missing Values in Visual Data
Liu, J., Musialski, P., Wonka, P., & Ye, J. (2013) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2012.39 · Cited by 2,063
Compressed sensing : theory and applications
Eldar, Y. C., & Kutyniok, G. (2012) · Cited by 2,030
Proceedings IEEE International Conference on Acoustics Speech and Signal Processing
Zheng, F., McLaughlin, S., & Mulgrew, B. (1991) · Cited by 1,905
Candes, E. J., & Plan, Y. (2010) · Proceedings of the IEEE · doi:10.1109/jproc.2009.2035722 · Cited by 1,709
Kruskal, J. B. (1977) · Linear Algebra and its Applications · doi:10.1016/0024-3795(77)90069-6 · Cited by 1,619
Tensor Decomposition for Signal Processing and Machine Learning
Sidiropoulos, N. D., De Lathauwer, L., Fu, X., Huang, K., Papalexakis, E. E., & Faloutsos, C. (2017) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2017.2690524 · Cited by 1,617 · cited 2×
“Readers interested in tensor decompositions and their applications could go further into [4, 106, 154, 174] for a more comprehensive introduction.”
BackgroundOn the Best Rank-1 and Rank-(R1 ,R2 ,. . .,RN) Approximation of Higher-Order Tensors
De Lathauwer, L., De Moor, B., & Vandewalle, J. (2000) · SIAM Journal on Matrix Analysis and Applications · doi:10.1137/s0895479898346995 · Cited by 1,567 · cited 1×
“is method was combined with the higher-order orthogonal iteration (HOOI) algorithm [36] to deal with missing values in paern-recognition problems in [55, 56].”
MethodCichocki, A., Zdunek, R., Phan, A. H., & Amari, S. (2009) · doi:10.1002/9780470747278 · Cited by 1,540
Bayesian probabilistic matrix factorization using Markov chain Monte Carlo
Salakhutdinov, R., & Mnih, A. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390267 · Cited by 1,413 · cited 1×
“They extend the matrix-based probabilistic approach as well as the MCMC optimization method proposed in [171] and demonstrates the advantage of a high-order temporal model over matrix-based static models.”
MethodThe Expression of a Tensor or a Polyadic as a Sum of Products
Hitchcock, F. L. (1927) · Journal of Mathematics and Physics · doi:10.1002/sapm192761164 · Cited by 1,345 · cited 4×
“CP Decomposition is first proposed by Hitchcock [64] and further discussed by Carroll [25] and Harshman [60].”
BackgroundMethodInfluentialAndersson, C. A., & Bro, R. (2000) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/s0169-7439(00)00071-x · Cited by 1,229
Relational learning via collective matrix factorization
Singh, A. P., & Gordon, G. J. (2008) · Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/1401890.1401969 · Cited by 1,204 · cited 2×
“wider margin which demonstrates the superiority of multidimensional analysis when more contextual information is available. Insomesituations, sideinformationofusers, itemsaswellasfeaturesareavailable[9,58,172,224], which motivates us to combine auxiliary information for collective learning. Borrowing the idea from mat…”
BackgroundMethodMulti‐Way Analysis with Applications in the Chemical Sciences
Smilde, A., Bro, R., & Geladi, P. (2004) · Wiley eBooks · doi:10.1002/0470012110 · Cited by 1,176
Mathematical Thought from Ancient to Modern Times
Kline, M. (1990) · doi:10.1093/oso/9780195061352.001.0001 · Cited by 1,171
Xu, Y., & Yin, W. (2013) · SIAM Journal on Imaging Sciences · doi:10.1137/120887795 · Cited by 1,160 · cited 3×
“Real-world applications sometimes require nonnegative constraints such as image completion [216], medical data analysis [80, 208] and so on.”
BackgroundMethodExact matrix completion via convex optimization
Candès, E., & Recht, B. (2012) · Communications of the ACM · doi:10.1145/2184319.2184343 · Cited by 996 · cited 6×
“In matrix case, Bernoulli sampling [26] or independent sampling with replacement [164] are always assumed to simply the random sampling approach.”
BackgroundInfluentialNovel Methods for Multilinear Data Completion and De-noising Based on Tensor-SVD
Zhang, Z., Ely, G., Aeron, S., Hao, N., & Kilmer, M. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.485 · Cited by 798 · cited 4×
“nducted by exploiting the sharp properties of the Gaussian measurement. Di‡erent from matrix trace-norm structure, some other trace-norm based structures are also proposed such as tensor-nuclear-norm [82, 172] and incoherent trace norm [170]. „ey either target new rank structure (tensor tubal-rank) or consider the inc…”
BackgroundMethodInfluentialTensor completion and low-n-rank tensor recovery via convex optimization
Gandy, S., Recht, B., & Yamada, I. (2011) · Inverse Problems · doi:10.1088/0266-5611/27/2/025010 · Cited by 738 · cited 5×
“nsors. As its particular matrix case [24], to avoid being an underdetermined and intractable problem, low rank is o›en a necessary hypothesis to restrict the degree of freedoms of the missing entries [57, 107, 125, 171]. Since a tensor has di‡erent types of rank de•nitions, to give a relatively general mathematical fo…”
BackgroundMethodInfluentialPrincipal Component Analysis of Three-Mode Data by Means of Alternating Least Squares Algorithms
Kroonenberg, P. M., & de Leeuw, J. (1980) · Psychometrika · doi:10.1007/bf02293599 · Cited by 724 · cited 1×
“ons from likelihood functions to handle missing values and conduct imputation. 3.1.2 Tucker Based Methods. Tucker decomposition is proposed by Tucker [158] and further developed by Kroonenberg et al. [92] and De Lathauwer et al. [35]. Given an Nth-order tensor X, its Tucker decomposition is de•ned as an approximation…”
MethodInfluentialKaratzoglou, A., Amatriain, X., Baltrunas, L., & Oliver, N. (2010) · Proceedings of the fourth ACM conference on Recommender systems · doi:10.1145/1864708.1864727 · Cited by 721
Large-Scale Parallel Collaborative Filtering for the Netflix Prize
Zhou, Y., Wilkinson, D., Schreiber, R., & Pan, R. (2008) · Lecture Notes in Computer Science · doi:10.1007/978-3-540-68880-8_32 · Cited by 700 · cited 1×
“A recent ALS approach for partially observed matrices [242] is extensible and parallelizable for the CP-based tensor completion, which updates each latent matrix An row by row as follows: An (i, :) = (B (n) i + λIR ) −1 C (n) i , (38)”
MethodPairwise interaction tensor factorization for personalized tag recommendation
Rendle, S., & Schmidt-Thieme, L. (2010) · Proceedings of the third ACM international conference on Web search and data mining · doi:10.1145/1718487.1718498 · Cited by 675 · cited 1×
“Thus, several metrics for classification problems such as precision (or precision@k), recall, F-measure and AUC score could be found in some tensor completion papers [60, 61, 156, 168, 192].”
BackgroundExact Tensor Completion Using t-SVD
Zhang, Z., & Aeron, S. (2016) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2016.2639466 · Cited by 675 · cited 2×
“aussian measurements N.A. O„rI2” Square Deal [138] Gaussian measurements N.A. O„r» 3 2 …I» 3 2 …” GoG [204] Uniformly Random Sample Incoherent O„ r 7 2 I 3 2 log 7 2 „I”+ 7Ilog6 ”” tubal-rank r t-SVD [219] Uniformly Random Sample Incoherent O„rI2log„I”” Table 3. Statistical assumptions & bounds (third-order tensor…”
BackgroundTensor Spaces and Numerical Tensor Calculus
Hackbusch, W. (2012) · Springer Series in Computational Mathematics · doi:10.1007/978-3-642-28027-6 · Cited by 607 · cited 1×
“Also called simple tensor [77] or decomposable tensor [72].”
BackgroundScalable tensor factorizations for incomplete data
Acar, E., Dunlavy, D. M., Kolda, T. G., & Mørup, M. (2011) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/j.chemolab.2010.08.004 · Cited by 602 · cited 9×
“[2] use multi-channel EEG (electroencephalogram) signals where missing data is encountered due to disconnections of electrodes and demonstrate the usefulness of scalable CP factorization method in capturing the underlying brain dynamics for incomplete data.”
BackgroundMethodInfluentialRecovering Low-Rank Matrices From Few Coefficients in Any Basis
Gross, D. (2011) · IEEE Transactions on Information Theory · doi:10.1109/tit.2011.2104999 · Cited by 602 · cited 1×
“In a matrix case, a given incoherentm × n matrix could be recovered with high probability if the sample size is larger than Cnr Û polyloд(n) for some constant C , where r is the matrix rank [59].”
BackgroundSocial Media Mining: An Introduction
Zafarani, R., Abbasi, M. A., & Liu, H. (2014) · Cambridge University Press eBooks · doi:10.1017/cbo9781139088510 · Cited by 592
Large-scale matrix factorization with distributed stochastic gradient descent
Gemulla, R., Nijkamp, E., Haas, P. J., & Sismanis, Y. (2011) · Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/2020408.2020426 · Cited by 589 · cited 1×
“Hence, each machine has to load all the fixed matrices into its memory as a scalability bottleneck noted in [63, 227].”
BackgroundBayesian CP Factorization of Incomplete Tensors with Automatic Rank Determination
Zhao, Q., Zhang, L., & Cichocki, A. (2015) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2015.2392756 · Cited by 587 · cited 1×
“From the perspective of probabilistic approaches [15, 117, 236, 237], these similarity matrices among each mode could all be treated as the inverse of kernel matrices in Gaussian prior distributions defined for each factor matrices.”
BackgroundHåstad, J. (1990) · Journal of Algorithms · doi:10.1016/0196-6774(90)90014-6 · Cited by 584
Recovering low-rank matrices from few coefficients in any basisPreprint
Gross, D. (2010) · Cited by 583
A literature survey of low‐rank tensor approximation techniques
Grasedyck, L., Kressner, D., & Tobler, C. (2013) · GAMM-Mitteilungen · doi:10.1002/gamm.201310004 · Cited by 577 · cited 1×
“Existing surveys are either resting on matrix level [72, 87, 96, 146] or focusing on related topics, e.g., tensor decompositions [4, 31, 46, 58, 62, 80, 84, 90, 91, 109, 121, 135, 141].”
BackgroundBig Data for Modern Industry: Challenges and Trends [Point of View]
Yin, S., & Kaynak, O. (2015) · Proceedings of the IEEE · doi:10.1109/jproc.2015.2388958 · Cited by 565
Duarte, M. F., & Baraniuk, R. G. (2012) · IEEE Transactions on Image Processing · doi:10.1109/tip.2011.2165289 · Cited by 546 · cited 2×
“In the area of computer vision, various problems could be formulated as tensor completion problems, such as image inpainting [59], video decoding [129], compressed sensing [46], and spectral data analysis [178].”
BackgroundMethodTravel time estimation of a path using sparse trajectories
Wang, Y., Zheng, Y., & Xue, Y. (2014) · Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/2623330.2623656 · Cited by 525 · cited 3×
“Deriving from coupled matrix factorization [179, 181], the way of coupled side information has also been widely developed in multi-way data analysis [3, 12, 16, 52, 127, 209, 239].”
BackgroundMethodBeyond streams and graphs: dynamic tensor analysis
Sun, J., Tao, D., & Faloutsos, C. (2006) · Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/1150402.1150445 · Cited by 462
Unsupervised Multiway Data Analysis: A Literature Survey
Acar, E., & Yener, B. (2009) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2008.112 · Cited by 434 · cited 1×
“Readers interested in tensor decompositions and their applications could go further into [4, 106, 154, 174] for a more comprehensive introduction.”
BackgroundPARAFAC2—Part I. A direct fitting algorithm for the PARAFAC2 model
Kiers, H. A. L., ten Berge, J. M. F., & Bro, R. (1999) · Journal of Chemometrics · doi:10.1002/(sici)1099-128x(199905/08)13:3/4<275::aid-cem543>3.0.co;2-b · Cited by 412 · cited 1×
“is approach is usually used in alternating projection optimizations and is also mentioned in some other early works for handling missing entries in PARAFAC model [14, 81, 90].”
MethodThree-mode principal component analysis : theory and applications
Kroonenberg, P. M. (1983) · Leiden Repository (Leiden University) · Cited by 410
Robust Low-Rank Tensor Recovery: Models and Algorithms
Goldfarb, D., & Qin, Z. (. (2014) · SIAM Journal on Matrix Analysis and Applications · doi:10.1137/130905010 · Cited by 408 · cited 5×
“First, as data in real-world applications are often corrupted with various of noises, different noise assumptions, especially Gaussian noises, are commonly used by researchers [14, 59, 67], leading to the following noise-contained formulation:”
BackgroundMethodInfluentialTemporal Link Prediction Using Matrix and Tensor Factorizations
Dunlavy, D. M., Kolda, T. G., & Acar, E. (2011) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/1921632.1921636 · Cited by 375 · cited 1×
“Their subsequent work in [47] gives another definition of CP score using the Holt-Winters forecasting method [32] and proves the effectiveness of tensor-based methods in forecasting the varying periodic patterns among links.”
MethodLearning optimal ranking with tensor factorization for tag recommendation
Rendle, S., Marinho, L. B., Νανόπουλος, Α., & Schmidt-Thieme, L. (2009) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/1557019.1557100 · Cited by 362
Scalable Tensor Decompositions for Multi-aspect Data Mining
Kolda, T. G., & Sun, J. (2008) · 2008 Eighth IEEE International Conference on Data Mining · doi:10.1109/icdm.2008.89 · Cited by 358 · cited 1×
“onal way to avoid intermediate data explosion in tensor completion is to reduce the operations that may be too large to •t in memory like the tensor-matrix multiplication. For instance, Kolda and Sun [104] work on the Tucker decomposition method for sparse data and solve the intermediate data explosion problem on the…”
MethodA survey of multilinear subspace learning for tensor data
Lu, H., Plataniotis, K. N., & Venetsanopoulos, A. N. (2011) · Pattern Recognition · doi:10.1016/j.patcog.2011.01.004 · Cited by 357 · cited 1×
“In one direction, tensor decomposition – which aims to distill tensors into (interpretable) soft representations – has been widely studied from both a theoretical and application-specific perspective [106, 135, 154, 174].”
BackgroundAdaptive signal processing algorithms: stability and performance
Solo, V., & Kong, X. (1994) · Prentice-Hall, Inc eBooks · Cited by 347
A weighted non-negative least squares algorithm for three-way ‘PARAFAC’ factor analysis
Paatero, P. (1997) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/s0169-7439(97)00031-2 · Cited by 340
An alternating direction algorithm for matrix completion with nonnegative factors
Xu, Y., Yin, W., Wen, Z., & Zhang, Y. (2012) · Frontiers of Mathematics in China · doi:10.1007/s11464-012-0194-5 · Cited by 339 · cited 1×
“For non-negative tensors with integer entries, Takeuchi and Ueda [193] use generalized Kullback–Leibler (gKL) divergence by analogy with several matrixbased approaches [21, 44, 217].”
MethodMultiple Invariants and Generalized Rank of a P‐Way Matrix or Tensor
Hitchcock, F. L. (1928) · Journal of Mathematics and Physics · doi:10.1002/sapm19287139 · Cited by 338
Holt-Winters Forecasting: Some Practical Issues
Chatfield, C., & Yar, M. (1988) · The Statistician · doi:10.2307/2348687 · Cited by 330 · cited 1×
“Their subsequent work in [47] gives another definition of CP score using the Holt-Winters forecasting method [32] and proves the effectiveness of tensor-based methods in forecasting the varying periodic patterns among links.”
MethodSrebro, N., & Shraibman, A. (2005) · Lecture Notes in Computer Science · doi:10.1007/11503415_37 · Cited by 316 · cited 1×
“In [187], for a matrix X with low-rank decomposition X = UV , an equivalent expression of its trace norm could be defined as”
BackgroundInductive matrix completion for predicting gene–disease associations
Natarajan, N., & Dhillon, I. S. (2014) · Bioinformatics · doi:10.1093/bioinformatics/btu269 · Cited by 311 · cited 1×
“It is an online inductive framework extended from the matrix completion framework proposed by [145, 173] and achieves superior performance in both batch and streaming settings.”
MethodTag recommendations based on tensor dimensionality reduction
Symeonidis, P., Nanopoulos, A., & Manolopoulos, Y. (2008) · Proceedings of the 2008 ACM conference on Recommender systems · doi:10.1145/1454008.1454017 · Cited by 310 · cited 2×
“Thus, several metrics for classification problems such as precision (or precision@k), recall, F-measure and AUC score could be found in some tensor completion papers [60, 61, 156, 168, 192].”
BackgroundLow-rank tensor completion by Riemannian optimization
Kressner, D., Steinlechner, M., & Vandereycken, B. (2014) · BIT Numerical Mathematics · doi:10.1007/s10543-013-0455-z · Cited by 290 · cited 9×
“Maps such as HOSVD method [110] and hierarchical singular value thresholding [161], which satisfy the necessary properties of a retraction mentioned in [110] could be used.”
BackgroundMethodInfluentialSmooth PARAFAC Decomposition for Tensor Completion
Yokota, T., Zhao, Q., & Cichocki, A. (2016) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2016.2586759 · Cited by 279 · cited 2×
“f regularization constraints based on the auxiliary information, in the end, we describe several other constraints leveraging di‡erent side information here for be−er coverage of related advances. In [214], the authors proposed a smooth constrained completion method, which imposes linear constraints on individual comp…”
BackgroundMethodIncremental tensor analysis: Theory and applications
Sun, J., Tao, D., Papadimitriou, S., Yu, P. S., & Faloutsos, C. (2008) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/1409620.1409621 · Cited by 254
Scalable Coordinate Descent Approaches to Parallel Matrix Factorization for Recommender Systems
Yu, H.-F., Hsieh, C.-J., Si, S., & Dhillon, I. (2012) · 2012 IEEE 12th International Conference on Data Mining · doi:10.1109/icdm.2012.168 · Cited by 253 · cited 3×
“on Problem. Prior research studies have put their main focus on the intermediate data explosion problem occurring in the operations for updating factor matrices in alternating least squares algorithm [8, 14, 33, 86, 165, 215]. With the increase of the size of input data, the size of the intermediate data in the operat…”
BackgroundMethodBayesian Robust Tensor Factorization for Incomplete Multiway Data
Zhao, Q., Zhou, G., Zhang, L., Cichocki, A. S., & Амари, Ш.-И. (2015) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2015.2423694 · Cited by 247 · cited 1×
“al and practical analysis of convex and non-convex models. Recent work of Jain [84] also adoptsADMMframeworktocompletenoisytensorswithoneoftheCPfactorsbeingsparse. Besides ADMM framework, Zhao et al. [223] consider the probabilistic framework with a fully Bayesian treatment optimized by variational Bayesian inference…”
MethodGigaTensor: scaling tensor analysis up by 100 times - algorithms and discoveries
Kang, U., Papalexakis, E., Harpale, A., & Faloutsos, C. (2012) · Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/2339530.2339583 · Cited by 238
On Tensor Completion via Nuclear Norm Minimization
Yuan, M., & Zhang, C.-H. (2016) · Foundations of Computational Mathematics · doi:10.1007/s10208-015-9269-5 · Cited by 236 · cited 1×
“Recent research has focused on minimizing the tensor CPrank directly via controlling it with tensor nuclear norm [218] and analyzing it theoretically [231], which provide more practical and general solutions for the LRTC problem under CP-rank constraints.”
All-at-once Optimization for Coupled Matrix and Tensor Factorizations
Acar, E., Kolda, T. G., & Dunlavy, D. (2011) · arXiv (Cornell University) · doi:10.48550/arxiv.1105.3422 · Cited by 233 · cited 3×
“[3] also show that though CP decomposition achieves comparable recovery accuracy when the missing ratio is lower than 80%, its recovery error sharply increased when the error rate is further increased.”
BackgroundMethodEstimation of low-rank tensors via convex optimizationPreprint
Tomioka, R., Hayashi, K., & Kashima, H. (2010) · arXiv (Cornell University) · doi:10.48550/arxiv.1010.0789 · Cited by 222 · cited 4×
“ches do. A simpli•ed version of it and two advanced algorithms were proposed in their subsequent journal paper [125]. A more popular approach for solving the above problem is to apply spli−ing method [57, 125, 169, 193], e.g., Alternating Direction Method of Multipliers (ADMM) which is a special case of Douglas-Rachfo…”
MethodInfluentialMetaFac: community discovery via relational hypergraph factorization
Lin, Y.-R., Sun, J., Castro, P., Konuru, R., Sundaram, H., & Kelliher, A. (2009) · Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/1557019.1557080 · Cited by 211
Strategies for analyzing data from video fluorometric monitoring of liquid chromatographic effluents
Appellof, C. J., & Davidson, E. R. (1981) · Analytical Chemistry · doi:10.1021/ac00236a025 · Cited by 205 · cited 2×
“Appellof and Davidson [6] are credited as the first one who applies tensor factorization in chemometrics [84].”
BackgroundLearning with tensors: a framework based on convex optimization and spectral regularization
Signoretto, M., Tran Dinh, Q., De Lathauwer, L., & Suykens, J. A. K. (2014) · Machine Learning · doi:10.1007/s10994-013-5366-3 · Cited by 204 · cited 1×
“[177] consider the problem of facial image completion using a benchmark”
MethodSquare Deal: Lower Bounds and Improved Relaxations for Tensor Recovery
Mu, C., Huang, B., Wright, J. N., & Goldfarb, D. (2013) · arXiv (Cornell University) · doi:10.48550/arxiv.1307.5870 · Cited by 203 · cited 5×
“N Block CD EM-like FaLRTC [125] SNN Smoothing Scheme Scheme EM-like HaLRTC [125] SNN ADMM EM-like [193] SNN ADMM EM-like (A/E)-CMLE [169] SNN ADMM EM-like ADM-TR (E) [57] SNN ADMM EM-like Square Deal [138] Single NN matrix ALM [123] EM-like [162] Alternative Relaxation ADMM EM-like [217] Incoherent Tensor Norms Gradie…”
BackgroundMethodInfluentialWalczak, B., & Massart, D. L. (2001) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/s0169-7439(01)00131-9 · Cited by 202
Adaptive Algorithms to Track the PARAFAC Decomposition of a Third-Order Tensor
Nion, D., & Sidiropoulos, N. D. (2009) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2009.2016885 · Cited by 201 · cited 1×
“omplexity. 6 DYNAMIC TENSOR COMPLETION Beyond traditional static/batch se−ing, with the increasing amount of high-velocity streaming data, more concerns have been addressed on dynamic tensor analysis [117, 147, 148]. As most of the existing surveys are lack of detailed introduction of dynamic tensor analysis, we •rst…”
BackgroundWeighted Least Squares Fitting Using Ordinary Least Squares Algorithms
Kiers, H. A. L. (1997) · Psychometrika · doi:10.1007/bf02295279 · Cited by 199 · cited 1×
“Borrowing the idea from weighted least square method applied to matrix case [78], Bro [15] demonstrates two ways of handling missing data in PARAFAC model, which might be one of the earliest approaches targeting to missing data imputation problem in multi-way data analysis.”
MethodLow rank tensor recovery via iterative hard thresholding
Rauhut, H., Schneider, R., & Stojanac, Ž. (2017) · Linear Algebra and its Applications · doi:10.1016/j.laa.2017.02.028 · Cited by 195
Computing Sparse Representations of Multidimensional Signals Using Kronecker Bases
Caiafa, C. F., & Cichocki, A. (2013) · Neural Computation · doi:10.1162/neco_a_00385 · Cited by 184 · cited 1×
“ompressedsensing,theKronecker-basedcompressedsensingmodel[44] has been proposed and explored in order to o‡er a practical implementation of compressed sensing for high-order tensors. Some researchers [21, 22, 118] focus on building generalized tensor compressed sensing algorithms by exploiting the Kronecker structure…”
MethodLink Prediction on Evolving Data Using Matrix and Tensor Factorizations
Acar, E., Dunlavy, D., & Kolda, T. G. (2009) · IEEE ... International Conference on Data Mining workshops · doi:10.1109/icdmw.2009.54 · Cited by 184
Simultaneous Tensor Decomposition and Completion Using Factor Priors
Chen, Y., Hsu, C.-T., & Liao, H.-Y. M. (2013) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2013.164 · Cited by 183 · cited 1×
“ge of the “within-mode regularization” and combine it with the coupled matrix method for the mobile recommendation problem. Similar idea of these two approaches have also been used in [8, 52, 53] and [27] respectively. From the perspective of probabilistic approaches [11, 95], these similarity matrices among each mode…”
MethodTensor completion using total variation and low-rank matrix factorization
Ji, T.-Y., Huang, T.-Z., Zhao, X.-L., Ma, T.-H., & Liu, G. (2016) · Information Sciences · doi:10.1016/j.ins.2015.07.049 · Cited by 173 · cited 1×
“There are still various of interesting methods for solving the completion problem such as alternative convex relaxation approach for Tucker ranks [169], adaptive sampling methods [18, 111] or various matrix-based methods [89].”
MethodTri-Clustered Tensor Completion for Social-Aware Image Tag Refinement
Tang, J., Shu, X., Qi, G.-J., Li, Z., Wang, M., Yan, S., & Jain, R. (2016) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2608882 · Cited by 170
Learning algorithms utilizing quasi-geodesic flows on the Stiefel manifold
Nishimori, Y., & Akaho, S. (2005) · Neurocomputing · doi:10.1016/j.neucom.2004.11.035 · Cited by 168
Tensor Versus Matrix Completion: A Comparison With Application to Spectral Data
Signoretto, M., de Plas, R. V., Moor, B. D., & Suykens, J. A. K. (2011) · IEEE Signal Processing Letters · doi:10.1109/lsp.2011.2151856 · Cited by 162 · cited 3×
“As its particular matrix case [24], to avoid being an underdetermined and intractable problem, low rank is oen a necessary hypothesis to restrict the degree of freedoms of the missing entries [57, 107, 125, 171].”
BackgroundInfluentialIncremental Tensor Subspace Learning and Its Applications to Foreground Segmentation and Tracking
Hu, W., Li, X., Zhang, X., Shi, X., Maybank, S. J., & Zhang, Z. M. (2010) · International Journal of Computer Vision · doi:10.1007/s11263-010-0399-6 · Cited by 161
Rubik: Knowledge Guided Tensor Factorization and Completion for Health Data Analytics
Wang, Y., Chen, R., Ghosh, J., Denny, J. C., Kho, A., Chen, Y., Malin, B. A., & Sun, J. (2015) · Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2783258.2783395 · Cited by 157
Information limits on neural identification of colored surfaces in natural scenes
FOSTER, D. H., Nascimento, S., & Amano, K. (2004) · Visual Neuroscience · doi:10.1017/s0952523804213335 · Cited by 157
HaTen2: Billion-scale tensor decompositions
Jeon, I., Papalexakis, E. E., Kang, U., & Faloutsos, C. (2015) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde.2015.7113355 · Cited by 153 · cited 6×
“All these operations are implemented in MAPREDUCE by designing mapper and reducer.”
MethodInfluentialLimestone: High-throughput candidate phenotype generation via tensor factorization
Ho, J. C., Ghosh, J., Steinhubl, S. R., Stewart, W. F., Denny, J. C., Malin, B. A., & Sun, J. (2014) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2014.07.001 · Cited by 151 · cited 1×
“Real-world applications sometimes require nonnegative constraints such as image completion [216], medical data analysis [80, 208] and so on.”
BackgroundTensor Spaces and Numerical Tensor Calculus
Hackbusch, W. (2019) · Springer series in computational mathematics · doi:10.1007/978-3-030-35554-8 · Cited by 148
Tucker factorization with missing data with application to low- $$n$$ n -rank tensor completion
Filipović, M., & Jukić, A. (2013) · Multidimensional Systems and Signal Processing · doi:10.1007/s11045-013-0269-9 · Cited by 148
Three-way component analysis: Principles and illustrative application.
Kiers, H. A. L., & Mechelen, I. V. (2001) · Psychological Methods · doi:10.1037/1082-989x.6.1.84 · Cited by 147
Link prediction in heterogeneous data via generalized coupled tensor factorization
Ermiş, B., Acar, E., & Cemgil, A. T. (2013) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-013-0341-y · Cited by 144 · cited 3×
“Deriving from coupled matrix factorization [179, 181], the way of coupled side information has also been widely developed in multi-way data analysis [3, 12, 16, 52, 127, 209, 239].”
MethodMulti-Way Compressed Sensing for Sparse Low-Rank Tensors
Sidiropoulos, N. D., & Kyrillidis, A. (2012) · IEEE Signal Processing Letters · doi:10.1109/lsp.2012.2210872 · Cited by 142 · cited 1×
“researchers [18, 19, 98] focus on building generalized tensor compressed sensing algorithms by exploiting the Kronecker structure with various approaches such as ‘ 1-minimization. Sidiropoulos et al. [136] recovers sparse, low-rank tensors from few measurements by leveraging compressed sensing for multilinear models o…”
MethodMulti-way Clustering on Relation Graphs
Banerjee, A., Basu, S., & Merugu, S. (2007) · SIAM International Conference on Data Mining · doi:10.1137/1.9781611972771.14 · Cited by 139 · cited 1×
“Deriving from coupled matrix factorization [179, 181], the way of coupled side information has also been widely developed in multi-way data analysis [3, 12, 16, 52, 127, 209, 239].”
MethodConditional gradient algorithms for norm-regularized smooth convex optimization
Harchaoui, Z., Juditsky, A., & Nemirovski, A. S. (2014) · Mathematical Programming · doi:10.1007/s10107-014-0778-9 · Cited by 133 · cited 1×
“By constructing a polytopal approximation of the tensor dual spectral norm, the completion problem regularized with tensor spectral norm could be solved by the generalized conditional gradient algorithm [71].”
MethodTrace Norm Regularized CANDECOMP/PARAFAC Decomposition With Missing Data
Liu, Y., Shang, F., Jiao, L., Cheng, J. S.-C., & Cheng, H. (2014) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2014.2374695 · Cited by 132 · cited 5×
“[108] also use Brain MRI dataset for testing their completion method which is a combination of trace-norm based and decomposition-based approach.”
BackgroundMethodInfluentialLink Propagation: A Fast Semi-supervised Learning Algorithm for Link Prediction
Kashima, H., Kato, T., Yamanishi, Y., Sugiyama, M., & Tsuda, K. (2009) · SIAM International Conference on Data Mining · doi:10.1137/1.9781611972795.94 · Cited by 130
Accelerating Online CP Decompositions for Higher Order Tensors
Zhou, S., Vinh, N. X., Bailey, J., Jia, Y., & Davidson, I. A. K. (2016) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/2939672.2939763 · Cited by 125 · cited 3×
“[240] develops an accelerated online CP algorithm that can incrementally track the decompositions of one-mode-change N th tensors.”
BackgroundMethodHankel Matrix Nuclear Norm Regularized Tensor Completion for $N$-dimensional Exponential Signals
Ying, J., Lu, H., Wei, Q., Cai, J., Guo, D., Wu, J., Chen, Z., & Qu, X. (2017) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2017.2695566 · Cited by 119 · cited 2×
“[167] focus on the exponential signal completion problem.”
BackgroundMissing Slice Recovery for Tensors Using a Low-Rank Model in Embedded Space
Yokota, T., Erem, B., Güler, Ş., et al. (2018) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2018.00861 · Cited by 118
Fast mining and forecasting of complex time-stamped events
Matsubara, Y., Sakurai, Y., Faloutsos, C., Iwata, T., & Yoshikawa, M. (2012) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/2339530.2339577 · Cited by 113 · cited 1×
“Matsubara [139] use topic modeling with Gibbs sampling method to get the posterior latent factors of each mode to solve the future traffic forecasting problem, e.”
MethodMultiway multiblock component and covariates regression models
Smilde, A. K., Westerhuis, J. A., & Boqué, R. (2000) · Journal of Chemometrics · doi:10.1002/1099-128x(200005/06)14:3<301::aid-cem594>3.0.co;2-h · Cited by 110 · cited 1×
“Factorization Another popular way of incorporating auxiliary information is to couple tensors or matrices together for jointly factorization and imputation. Deriving from coupled matrix factorization [172, 174], the way of coupled side information has also been widely developed in multi-way data analysis [3, 10, 14, 5…”
MethodRank Regularization and Bayesian Inference for Tensor Completion and Extrapolation
Bazerque, J. A., Mateos, G., & Giannakis, G. B. (2013) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2013.2278516 · Cited by 100 · cited 4×
“[15] perform experiments on RNA sequencing data which represents the gene expression levels in yeast.”
BackgroundMethodInfluentialOptimization on the Hierarchical Tucker manifold – Applications to tensor completion
Silva, C. D., & Herrmann, F. J. (2015) · Linear Algebra and its Applications · doi:10.1016/j.laa.2015.04.015 · Cited by 94 · cited 1×
“Rauhut [154] reconstructs a low-rank tensor from a relatively small number of measurements by utilizing the framework of hierarchical tensor (HT) formats based on a Hierarchical SVD-based approach followed by a similar work [38].”
MethodCompressive Sensing of Sparse Tensors
Friedland, S., Li, Q., & Schonfeld, D. (2014) · IEEE Transactions on Image Processing · doi:10.1109/tip.2014.2348796 · Cited by 86 · cited 1×
“ollowed by a similar work [38]. However, it should be noted that this method is hard to scale tensors with large mode sizes as it strongly relies on computing SVDs of large matrices. Friedland et al. [56] exploit a uni•ed framework of compressed sensing for high-order tensors with preserving the intrinsic structure of…”
MethodHigh-order tensor completion via gradient-based optimization under tensor train format
Yuan, L., Zhao, Q., Gui, L., & Cao, J. (2018) · Signal Processing Image Communication · doi:10.1016/j.image.2018.11.012 · Cited by 86
Variants of Alternating Least Squares Tensor Completion in the Tensor Train Format
Grasedyck, L., Kluge, M., & Krämer, S. (2015) · SIAM Journal on Scientific Computing · doi:10.1137/130942401 · Cited by 83
Fully Scalable Methods for Distributed Tensor Factorization
Shin, K., Sael, L., & Kang, U. (2016) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2016.2610420 · Cited by 77 · cited 9×
“(right) [172] on a third-order tensor with four machines.”
BackgroundMethodInfluentialMulti-Aspect Streaming Tensor Completion
Song, Q., Huang, X. X., Ge, H., Caverlee, J., & Hu, X. (2017) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3097983.3098007 · Cited by 71 · cited 4×
“As shown in the left figure in Figure 8, different from the streaming tensor setting, in the multi-aspect streaming setting, a tensor could increase on any of its modes, which imposes greater challenge in coordinating multi-dimensional dynamics, handling the higher time and space complexity as well as imputing the inc…”
BackgroundMethodInfluentialSelf-citationOn Polynomial Time Methods for Exact Low-Rank Tensor Completion
Xia, D., & Yuan, M. (2019) · Foundations of Computational Mathematics · doi:10.1007/s10208-018-09408-6 · Cited by 70 · cited 1×
“GoG [211] Uniformly Random Sample Incoherent O(r 7 2 I 3 2 loд 7 2 (I ) + r 7I log6 (I ))”
Face Image Modeling by Multilinear Subspace Analysis With Missing Values
Geng, X., Smith‐Miles, K., Zhou, Z.-H., & Wang, L. (2011) · IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics) · doi:10.1109/tsmcb.2010.2097588 · Cited by 70
Collaborative filtering in social tagging systems based on joint item-tag recommendations
Peng, J., Zeng, D. D., Zhao, H., & Wang, F. (2010) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/1871437.1871541 · Cited by 69 · cited 2×
“Thus, several metrics for classification problems such as precision (or precision@k), recall, F-measure and AUC score could be found in some tensor completion papers [60, 61, 156, 168, 192].”
BackgroundSpectral Algorithms for Tensor Completion
Andrea Montanari, A., & Sun, N. (2018) · Communications on Pure and Applied Mathematics · doi:10.1002/cpa.21748 · Cited by 67 · cited 1×
“[142] with matching statistical guarantee (i.”
Bayesian Sparse Tucker Models for Dimension Reduction and Tensor CompletionPreprint
Zhao, Q., Zhang, L., & Cichocki, A. S. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1505.02343 · Cited by 61 · cited 1×
“From the perspective of probabilistic approaches [15, 117, 236, 237], these similarity matrices among each mode could all be treated as the inverse of kernel matrices in Gaussian prior distributions defined for each factor matrices.”
BackgroundSobral, A. C., Baker, C. G., Bouwmans, T., & Zahzah, E.-H. (2014) · Lecture notes in computer science · doi:10.1007/978-3-319-11758-4_11 · Cited by 60 · cited 1×
“Several other online Tucker decomposition methods are conducted recently not only targeting on the one-mode-increase pattern [81, 184] but also deriving possible solutions for multi-aspect streaming patterns drawing upon matrix-based online-learning methods such as incremental SVD [136].”
MethodIncoherent Tensor Norms and Their Applications in Higher Order Tensor Completion
Yuan, M., & Zhang, C. (2017) · IEEE Transactions on Information Theory · doi:10.1109/tit.2017.2724549 · Cited by 58 · cited 5×
“[230] Incoherent Tensor Norms Gradient based MS”
BackgroundInfluentialModel-Driven Sparse CP Decomposition for Higher-Order Tensors
Li, J., Choi, J. W., Perros, I., Sun, J., & Vuduc, R. W. (2017) · IEEE International Parallel and Distributed Processing Symposium (IPDPS) · doi:10.1109/ipdps.2017.80 · Cited by 57 · cited 1×
“CT and CDTF/SALS are compared in Figure6. Besides the aforementioned approaches, there are also other scalable tensor completion or sparse tensor decomposition approaches focusing on either CP format [60, 91, 96, 117, 149, 153, 175] or Tucker format [128, 144, 176]. As the end of this subsection, we compare several th…”
MethodAccelerating the Tucker Decomposition with Compressed Sparse Tensors
Smith, S., & Karypis, G. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-64203-1_47 · Cited by 57 · cited 1×
“Besides the aforementioned approaches, there are also other scalable tensor completion or sparse tensor decomposition approaches focusing on either CP format [62, 93, 98, 122, 155, 160, 182] or Tucker format [133, 150, 183].”
MethodGoal-Directed Inductive Matrix Completion
Si, S., Chiang, K.-Y., Hsieh, C., Rao, N., & Dhillon, I. S. (2016) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/2939672.2939809 · Cited by 56 · cited 1×
“It is an online inductive framework extended from the matrix completion framework proposed by [145, 173] and achieves superior performance in both batch and streaming settings.”
MethodTAPER: A Contextual Tensor-Based Approach for Personalized Expert Recommendation
Ge, H., Caverlee, J., & Lu, H. (2016) · ACM Conference on Recommender Systems (RecSys) · doi:10.1145/2959100.2959151 · Cited by 51
Self-citationR3MC: A Riemannian three-factor algorithm for low-rank matrix completion
Mishra, B., & Sepulchre, R. J. (2014) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc.2014.7039534 · Cited by 51 · cited 1×
“iagonal approximation of its Hessian and de•ned a novel metric on the tangent space of M r and its quotient manifold. „ey further generalized a preconditioned conjugate gradient algorithm proposed in [136] with a total computational cost O„jΩjr 1r 2r 3”for a three-order tensor. Finally, the retraction map is used map…”
MethodSimple tensor productsPreprint
Hernandez, D. (2013) · Cited by 46
Li, J., Bien, J., & Wells, M. T. (2018) · Journal of Statistical Software · doi:10.18637/jss.v087.i10 · Cited by 43
Leibovici, D. (2010) · Journal of Statistical Software · doi:10.18637/jss.v034.i10 · Cited by 42
Facial age estimation by multilinear subspace analysis
Geng, X., & Smith‐Miles, K. (2009) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2009.4959721 · Cited by 40 · cited 2×
“simultaneously complete the images and a label matrix of images. „eir spectral regularized framework tackles both pure completion problems and multi-task cases. Earlier works conducted by Geng et al. [62, 63] consider the “hard completion problem”. „ey leverage the HOOI method to model face images for face completion,…”
MethodMatrix and Tensor Decomposition in Recommender Systems
Symeonidis, P. (2016) · ACM Conference on Recommender Systems (RecSys) · doi:10.1145/2959100.2959195 · Cited by 40 · cited 1×
“, Social Tagging Systems (STSs)), many recommendation algorithms, which are originally designed to operate on matrices, cannot be easily applied on these multi-dimensional datasets [191].”
BackgroundHernandez, D. (2010) · Inventiones mathematicae · doi:10.1007/s00222-010-0256-9 · Cited by 39 · cited 1×
“Also called simple tensor [77] or decomposable tensor [72].”
BackgroundA New Sampling Technique for TensorsPreprint
Bhojanapalli, S., & Sanghavi, S. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1502.05023 · Cited by 37 · cited 1×
“based on hierarchical Tucker representations have been utilized to address high-order problems [124], their generalizability and robustness remain unclear. „ird, some methods such as sampling methods [13, 89] require speci•c assumptions (e.g., incoherence assumptions), which are hard to verify and guarantee for real-w…”
MethodA Rank-One Tensor Updating Algorithm for Tensor Completion
Yang, Y., Feng, Y., & Suykens, J. A. K. (2015) · IEEE Signal Processing Letters · doi:10.1109/lsp.2015.2420592 · Cited by 36 · cited 1×
“Recent research has focused on minimizing the tensor CPrank directly via controlling it with tensor nuclear norm [218] and analyzing it theoretically [231], which provide more practical and general solutions for the LRTC problem under CP-rank constraints.”
Stochastic Decomposition into Low Rank and Sparse Tensor for Robust Background Subtraction
Javed, S., Bouwmans, T., & Jung, S. K. (2015) · International Conference on Imaging for Crime Detection and Prevention (ICDP) · doi:10.1049/ic.2015.0105 · Cited by 34 · cited 1×
“[87] provide an online stochastic optimization method for robust low-rank and sparse error separation in addressing background subtraction problem.”
MethodSimultaneous Tensor Completion and Denoising by Noise Inequality Constrained Convex Optimization
Yokota, T., & Hontani, H. (2019) · IEEE Access · doi:10.1109/access.2019.2894622 · Cited by 32
Link Prediction via Generalized Coupled Tensor FactorisationPreprint
Ermiş, B., Acar, E., & Cemgil, A. T. (2012) · arXiv (Cornell University) · doi:10.48550/arxiv.1208.6231 · Cited by 29
Uncovering the Spatio-Temporal Dynamics of Memes in the Presence of Incomplete Information
Ge, H., Caverlee, J., Zhang, N., & Squicciarini, A. (2016) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/2983323.2983782 · Cited by 28 · cited 8×
“It is critical, yet challenging, to consider heterogeneous data sources to benefit the tensor completion problem, especially for structured missing data [61, 222].”
BackgroundResultInfluentialSelf-citationConstrained Tensor Factorization with Accelerated AO-ADMM
Smith, S., Beri, A., & Karypis, G. (2017) · Proceedings of the International Conference on Parallel Processing · doi:10.1109/icpp.2017.20 · Cited by 28 · cited 1×
“CT and CDTF/SALS are compared in Figure6. Besides the aforementioned approaches, there are also other scalable tensor completion or sparse tensor decomposition approaches focusing on either CP format [60, 91, 96, 117, 149, 153, 175] or Tucker format [128, 144, 176]. As the end of this subsection, we compare several th…”
MethodSael, L., Jeon, I., & Kang, U. (2015) · Big Data Research · doi:10.1016/j.bdr.2015.01.004 · Cited by 23 · cited 1×
“The incremental interest in big data analytics [91, 107, 170] motivates us to provide a view from big data perspective in the present survey that differs from most of the existing reviews.”
ResultField-of-Experts Filters Guided Tensor Completion
Xiong, B., Liu, Q., Xiong, J., Li, S., Wang, S., & Liang, D. (2018) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2018.2806225 · Cited by 20 · cited 1×
“ld data-driven applications, besides the target tensor object, additional side information such as spatial and temporal similarities among objects or auxiliary coupled matrices/tensors may also exist [58, 59, 139, 200, 206]. „ese heterogeneous data sources are usually bonded and compensate with each other and could se…”
BackgroundAutomatic image tagging and recommendation via PARAFAC2
Pantraki, E., & Kotropoulos, C. L. (2015) · IEEE International Workshop on Machine Learning for Signal Processing (MLSP) · doi:10.1109/mlsp.2015.7324363 · Cited by 15 · cited 1×
“, Parafac2 model [153]) or optimization techniques (e.”
MethodDisTenC: A Distributed Algorithm for Scalable Tensor Completion on Spark
Ge, H., Zhang, K., Alfifi, M., Hu, X., & Caverlee, J. (2018) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde.2018.00022 · Cited by 14 · cited 1×
“CT and CDTF/SALS are compared in Figure6. Besides the aforementioned approaches, there are also other scalable tensor completion or sparse tensor decomposition approaches focusing on either CP format [60, 91, 96, 117, 149, 153, 175] or Tucker format [128, 144, 176]. As the end of this subsection, we compare several th…”
MethodSelf-citationConvex Coupled Matrix and Tensor Completion
Wimalawarne, K., Yamada, M., & Mamitsuka, H. (2018) · Neural Computation · doi:10.1162/neco_a_01123 · Cited by 13
tensorBF: an R package for Bayesian tensor factorizationPreprint
Khan, S. A., & Ammad-ud-din, M. (2016) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/097048 · Cited by 10
Scalable and Sound Low-Rank Tensor Learning
Cheng, H., Yu, Y., Xin-hua, Z., Xing, E. P., & Schuurmans, D. (2016) · International Conference on Artificial Intelligence and Statistics · Cited by 8
Handling missing data in medical questionnaires using tensor decompositions
Dauwels, J., Garg, L., Earnest, A., & Pang, L. K. (2011) · International Conference on Information, Communications & Signal Processing (ICICS) · doi:10.1109/icics.2011.6174300 · Cited by 8 · cited 1×
“st online subspace tracking. Besides the applications on medical images, several other types of datasets and analytics are also conducted in the realm of bioinformatics and healthcare. Dauwels et al. [39] apply CP decomposition to handle missing data in medical questionnaires. Bazerque et al. [13] perform experiments…”
Background
Metadata
- Authors
- Song, Q., Ge, H., Caverlee, J., & Hu, X.
- Venue
- ACM Transactions on Knowledge Discovery from Data
- Published
- 2019-01-09
- Volume
- 13
- Issue
- 1
- Pages
- 1-48
- ISSN
- 1556-4681 (ISSN-L)
- 1556-472X
- OpenAlex
- W2963472624
- Cited by
- 229
- References
- 154
- FWCI
- 10.74
- Type
- journal article
- Language
- en
- Open access
- bronze
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.