Abstract

With many applications relying on multi-dimensional datasets for decision making, tensors (or multi-dimensional arrays) are emerging as a popular data representation to support diverse types of data, such as sensor streams and social networks. Consequently, tensor decomposition forms the basis for many data analysis and knowledge discovery tasks, from clustering, trend detection, anomaly detection, to correlation analysis. In applications where data evolves over time and the tensor-based analysis results need to be continuously maintained, re-computation of the whole tensor decomposition with each update will cause high computational costs and incur large memory overheads. In this paper, we propose a two-phase block-incremental CP-based tensor decomposition technique, BICP, that efficiently and effectively maintains tensor decomposition results in the presence of dynamically evolving tensor data. In its first phase, instead of repeatedly conducting ALS on each subtensor, BICP only revises the decompositions of the tensors that contain updated data. Moreover, when updates are relatively small with respect to the block size, BICP relies on a incremental factor tracking to avoid re-decomposition the updated sub-tensor. In its second phase, BICP limits the block-centric refinement process to only those blocks that are critical given the update. Experiment results show that the proposed method significantly reduces the execution time while assuring high accuracy.

Original languageEnglish (US)
Title of host publicationCIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management
PublisherAssociation for Computing Machinery
Pages1221-1230
Number of pages10
Volume24-28-October-2016
ISBN (Electronic)9781450340731
DOIs
StatePublished - Oct 24 2016
Event25th ACM International Conference on Information and Knowledge Management, CIKM 2016 - Indianapolis, United States
Duration: Oct 24 2016Oct 28 2016

Other

Other25th ACM International Conference on Information and Knowledge Management, CIKM 2016
CountryUnited States
CityIndianapolis
Period10/24/1610/28/16

Fingerprint

Incremental
Decomposition
Anomaly detection
Clustering
Factors
Knowledge discovery
Sensor
Correlation analysis
Experiment
Social networks
Costs
Decision making

ASJC Scopus subject areas

  • Business, Management and Accounting(all)
  • Decision Sciences(all)

Cite this

Huang, S., Candan, K., & Sapino, M. L. (2016). BICP: Block-incremental CP decomposition with update sensitive refinement. In CIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management (Vol. 24-28-October-2016, pp. 1221-1230). Association for Computing Machinery. https://doi.org/10.1145/2983323.2983717

BICP : Block-incremental CP decomposition with update sensitive refinement. / Huang, Shengyu; Candan, Kasim; Sapino, Maria Luisa.

CIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management. Vol. 24-28-October-2016 Association for Computing Machinery, 2016. p. 1221-1230.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Huang, S, Candan, K & Sapino, ML 2016, BICP: Block-incremental CP decomposition with update sensitive refinement. in CIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management. vol. 24-28-October-2016, Association for Computing Machinery, pp. 1221-1230, 25th ACM International Conference on Information and Knowledge Management, CIKM 2016, Indianapolis, United States, 10/24/16. https://doi.org/10.1145/2983323.2983717
Huang S, Candan K, Sapino ML. BICP: Block-incremental CP decomposition with update sensitive refinement. In CIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management. Vol. 24-28-October-2016. Association for Computing Machinery. 2016. p. 1221-1230 https://doi.org/10.1145/2983323.2983717
Huang, Shengyu ; Candan, Kasim ; Sapino, Maria Luisa. / BICP : Block-incremental CP decomposition with update sensitive refinement. CIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management. Vol. 24-28-October-2016 Association for Computing Machinery, 2016. pp. 1221-1230
@inproceedings{b5c26abd9ff249ceb31dd7197989d73c,
title = "BICP: Block-incremental CP decomposition with update sensitive refinement",
abstract = "With many applications relying on multi-dimensional datasets for decision making, tensors (or multi-dimensional arrays) are emerging as a popular data representation to support diverse types of data, such as sensor streams and social networks. Consequently, tensor decomposition forms the basis for many data analysis and knowledge discovery tasks, from clustering, trend detection, anomaly detection, to correlation analysis. In applications where data evolves over time and the tensor-based analysis results need to be continuously maintained, re-computation of the whole tensor decomposition with each update will cause high computational costs and incur large memory overheads. In this paper, we propose a two-phase block-incremental CP-based tensor decomposition technique, BICP, that efficiently and effectively maintains tensor decomposition results in the presence of dynamically evolving tensor data. In its first phase, instead of repeatedly conducting ALS on each subtensor, BICP only revises the decompositions of the tensors that contain updated data. Moreover, when updates are relatively small with respect to the block size, BICP relies on a incremental factor tracking to avoid re-decomposition the updated sub-tensor. In its second phase, BICP limits the block-centric refinement process to only those blocks that are critical given the update. Experiment results show that the proposed method significantly reduces the execution time while assuring high accuracy.",
author = "Shengyu Huang and Kasim Candan and Sapino, {Maria Luisa}",
year = "2016",
month = "10",
day = "24",
doi = "10.1145/2983323.2983717",
language = "English (US)",
volume = "24-28-October-2016",
pages = "1221--1230",
booktitle = "CIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management",
publisher = "Association for Computing Machinery",

}

TY - GEN

T1 - BICP

T2 - Block-incremental CP decomposition with update sensitive refinement

AU - Huang, Shengyu

AU - Candan, Kasim

AU - Sapino, Maria Luisa

PY - 2016/10/24

Y1 - 2016/10/24

N2 - With many applications relying on multi-dimensional datasets for decision making, tensors (or multi-dimensional arrays) are emerging as a popular data representation to support diverse types of data, such as sensor streams and social networks. Consequently, tensor decomposition forms the basis for many data analysis and knowledge discovery tasks, from clustering, trend detection, anomaly detection, to correlation analysis. In applications where data evolves over time and the tensor-based analysis results need to be continuously maintained, re-computation of the whole tensor decomposition with each update will cause high computational costs and incur large memory overheads. In this paper, we propose a two-phase block-incremental CP-based tensor decomposition technique, BICP, that efficiently and effectively maintains tensor decomposition results in the presence of dynamically evolving tensor data. In its first phase, instead of repeatedly conducting ALS on each subtensor, BICP only revises the decompositions of the tensors that contain updated data. Moreover, when updates are relatively small with respect to the block size, BICP relies on a incremental factor tracking to avoid re-decomposition the updated sub-tensor. In its second phase, BICP limits the block-centric refinement process to only those blocks that are critical given the update. Experiment results show that the proposed method significantly reduces the execution time while assuring high accuracy.

AB - With many applications relying on multi-dimensional datasets for decision making, tensors (or multi-dimensional arrays) are emerging as a popular data representation to support diverse types of data, such as sensor streams and social networks. Consequently, tensor decomposition forms the basis for many data analysis and knowledge discovery tasks, from clustering, trend detection, anomaly detection, to correlation analysis. In applications where data evolves over time and the tensor-based analysis results need to be continuously maintained, re-computation of the whole tensor decomposition with each update will cause high computational costs and incur large memory overheads. In this paper, we propose a two-phase block-incremental CP-based tensor decomposition technique, BICP, that efficiently and effectively maintains tensor decomposition results in the presence of dynamically evolving tensor data. In its first phase, instead of repeatedly conducting ALS on each subtensor, BICP only revises the decompositions of the tensors that contain updated data. Moreover, when updates are relatively small with respect to the block size, BICP relies on a incremental factor tracking to avoid re-decomposition the updated sub-tensor. In its second phase, BICP limits the block-centric refinement process to only those blocks that are critical given the update. Experiment results show that the proposed method significantly reduces the execution time while assuring high accuracy.

UR - http://www.scopus.com/inward/record.url?scp=84996538084&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84996538084&partnerID=8YFLogxK

U2 - 10.1145/2983323.2983717

DO - 10.1145/2983323.2983717

M3 - Conference contribution

VL - 24-28-October-2016

SP - 1221

EP - 1230

BT - CIKM 2016 - Proceedings of the 2016 ACM Conference on Information and Knowledge Management

PB - Association for Computing Machinery

ER -