TY - GEN

T1 - IDR/QR

T2 - KDD-2004 - Proceedings of the Tenth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

AU - Ye, Jieping

AU - Li, Qi

AU - Xiong, Hui

AU - Park, Haesun

AU - Janardan, Ravi

AU - Kumar, Vipin

N1 - Copyright:
Copyright 2008 Elsevier B.V., All rights reserved.

PY - 2004

Y1 - 2004

N2 - Dimension reduction is critical for many database and data mining applications, such as efficient storage and retrieval of high-dimensional data. In the literature, a well-known dimension reduction scheme is Linear Discriminant Analysis (LDA). The common aspect of previously proposed LDA based algorithms is the use of Singular Value Decomposition (SVD). Due to the difficulty of designing an incremental solution for the eigenvalue problem on the product of scatter matrices in LDA, there is little work on designing incremental LDA algorithms. In this paper, we propose an LDA based incremental dimension reduction algorithm, called IDR/QR, which applies QR Decomposition rather than SVD. Unlike other LDA based algorithms, this algorithm does not require the whole data matrix in main memory. This is desirable for large data sets. More importantly, with the insertion of new data items, the IDR/QR algorithm can constrain the computational cost by applying efficient QR-updating techniques. Finally, we evaluate the effectiveness of the IDR/QR algorithm in terms of classification accuracy on the reduced dimensional space. Our experiments on several real-world data sets reveal that the accuracy achieved by the IDR/QR algorithm is very close to the best possible accuracy achieved by other LDA based algorithms. However, the IDR/QR algorithm has much less computational cost, especially when new data items are dynamically inserted.

AB - Dimension reduction is critical for many database and data mining applications, such as efficient storage and retrieval of high-dimensional data. In the literature, a well-known dimension reduction scheme is Linear Discriminant Analysis (LDA). The common aspect of previously proposed LDA based algorithms is the use of Singular Value Decomposition (SVD). Due to the difficulty of designing an incremental solution for the eigenvalue problem on the product of scatter matrices in LDA, there is little work on designing incremental LDA algorithms. In this paper, we propose an LDA based incremental dimension reduction algorithm, called IDR/QR, which applies QR Decomposition rather than SVD. Unlike other LDA based algorithms, this algorithm does not require the whole data matrix in main memory. This is desirable for large data sets. More importantly, with the insertion of new data items, the IDR/QR algorithm can constrain the computational cost by applying efficient QR-updating techniques. Finally, we evaluate the effectiveness of the IDR/QR algorithm in terms of classification accuracy on the reduced dimensional space. Our experiments on several real-world data sets reveal that the accuracy achieved by the IDR/QR algorithm is very close to the best possible accuracy achieved by other LDA based algorithms. However, the IDR/QR algorithm has much less computational cost, especially when new data items are dynamically inserted.

KW - Dimension reduction

KW - Incremental learning

KW - Linear Discriminant Analysis

KW - QR Decomposition

UR - http://www.scopus.com/inward/record.url?scp=12244268926&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=12244268926&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:12244268926

SN - 1581138881

SN - 9781581138887

T3 - KDD-2004 - Proceedings of the Tenth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

SP - 364

EP - 373

BT - KDD-2004 - Proceedings of the Tenth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

A2 - Kohavi, R.

A2 - Gehrke, J.

A2 - DuMouchel, W.

A2 - Ghosh, J.

Y2 - 22 August 2004 through 25 August 2004

ER -