Scalable Tensor Decompositions for Multi-aspect Data Mining

Kolda, Tamara G.; Sun, Jimeng

doi:10.1109/icdm.2008.89

Cited by 302 publications

(212 citation statements)

References 14 publications

Supporting

Mentioning

209

Contrasting

Unclassified

Order By: Relevance

“…Indeed,F is a dense matrix and even the scalable tensor decomposition technique proposed in [16] becomes unusable because it is designed for sparse tensors. The huge amount of computations and storage required forF would render the whole method useless.…”

Section: Cubelsi Algorithmmentioning

confidence: 99%

“…The analogy between corresponding properties of SVD and Tucker decomposition, e.g., uniqueness, have been investigated in depth. An efficient algorithm for Tucker decomposition, called MET, was proposed in [16]. MET works well with sparse high-dimensional data with the assumption that there is enough memory to fit the output tensor, which serves to compute pairwise tag distances, resulting from the decomposition.…”

Section: Related Workmentioning

confidence: 99%

“…Recall that the basic idea of CubeLSI is to apply Tucker decomposition on the third-order tensor F to obtain an output tensorF, (2) 8.8 MB 3.0 MB 1.8 MB based on which tag distances are computed. We have explained in Section 4.4 that, for a typical social tagging system, the tensorF is so huge that not even the state-of-the-art decomposition algorithm (such as MET [16]) can handle. Again, our two theorems allow CubeLSI to compute tag distances from two relatively small structures S and Y (2) without ever materializingF.…”

Section: Efficiencymentioning

confidence: 99%

See 2 more Smart Citations

CubeLSI: An effective and efficient method for searching resources in social tagging systems

Lee

Kao

et al. 2011

2011 IEEE 27th International Conference on Data Engineering

View full text Add to dashboard Cite

In a social tagging system, resources (such as photos, video and web pages) are associated with tags. These tags allow the resources to be effectively searched through tag-based keyword matching using traditional IR techniques. We note that in many such systems, tags of a resource are often assigned by a diverse audience of causal users (taggers). This leads to two issues that gravely affect the effectiveness of resource retrieval: (1) Noise: tags are picked from an uncontrolled vocabulary and are assigned by untrained taggers. The tags are thus noisy features in resource retrieval. (2) A multitude of aspects: different taggers focus on different aspects of a resource. Representing a resource using a flattened bag of tags ignores this important diversity of taggers. To improve the effectiveness of resource retrieval in social tagging systems, we propose CubeLSI -a technique that extends traditional LSI to include taggers as another dimension of feature space of resources. We compare CubeLSI against a number of other tag-based retrieval models and show that CubeLSI significantly outperforms the other models in terms of retrieval accuracy. We also prove two interesting theorems that allow CubeLSI to be very efficiently computed despite the much enlarged feature space it employs.

show abstract

Section: Cubelsi Algorithmmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Efficiencymentioning

confidence: 99%

See 1 more Smart Citation

CubeLSI: An effective and efficient method for searching resources in social tagging systems

Lee

Kao

et al. 2011

2011 IEEE 27th International Conference on Data Engineering

View full text Add to dashboard Cite

show abstract

“…[18] proposed a memory-efficient Tucker (MET) decomposition to address the intermediate blowup problem in Tucker decomposition by updating a subset of the modes at a time. [29] proposed MACH, a randomized algorithm (based on randomized sampling) that speedups the Tucker decomposition while providing accuracy guarantees.…”

Section: Pcp and Re-use Promoting Scheduling Of Block Accessesmentioning

confidence: 99%

“…Thanks to the availability of various mathematical tools (such as decompositions) that support multi-aspect analysis of data, tensors are increasingly being used for representing multi-dimensional data, such as sensor streams and social networks [29], [18], [14], [17], [19]. Matrix-shaped data (i.e., 2-mode tensors) are often analyzed for their latent semantics through matrix decomposition operations, such as singular value decomposition (SVD).…”

Section: Introductionmentioning

confidence: 99%

2PCP: Two-phase CP decomposition for billion-scale dense tensors

Huang

Candan

et al. 2016

2016 IEEE 32nd International Conference on Data Engineering (ICDE)

View full text Add to dashboard Cite

Abstract-Tensors are multi-dimensional arrays -consequently, tensor decomposition operations (CP and Tucker) are the bases for many high-dimensional data analysis tasks, from clustering, trend detection, anomaly detection, to correlation analysis in various application domains, including science and engineering 1 . One key problem with tensor decomposition is its computational complexity and space requirements. Especially, as the relevant data sets get denser, in-memory schemes for tensor decomposition become increasingly ineffective; therefore out-of-core (secondary-memory supported, potentially parallel) computing is necessitated. However, existing techniques do not consider the I/O and network data exchange costs that out-ofcore execution of the tensor decomposition operation will incur. In this paper, we note that when this operation is implemented with the help of secondary-memory and/or multiple servers to tackle the memory limitations, we would need intelligent buffer-management and task-scheduling techniques which take into account the cost of bringing the relevant blocks into the buffer to minimize I/O in the system. In this paper, we introduce 2PCP, a two-phase, block-based CP decomposition system with intelligent buffer sensitive task scheduling and buffer management mechanisms. 2PCP aims to reduce I/O costs in the analysis of relatively dense tensors common in scientific and engineering applications. Experiment results compare with current state of art tensor decomposition algorithms and show that our algorithms can significantly reduce the amount of I/O and execution time while maintaining decomposition accuracy.

show abstract