東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Efficient algorithms for sparse sing...

Rajamanickam, Sivasankaran.

FindBook

Google Book

Amazon

博客來

Efficient algorithms for sparse singular value decomposition .

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Efficient algorithms for sparse singular value decomposition ./
作者:	Rajamanickam, Sivasankaran.
面頁冊數:	97 p.
附註:	Source: Dissertation Abstracts International, Volume: 71-03, Section: B, page: 1821.
Contained By:	Dissertation Abstracts International71-03B.
標題:	Applied Mathematics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3400303
ISBN:	9781109670950

Efficient algorithms for sparse singular value decomposition .
Rajamanickam, Sivasankaran.

Efficient algorithms for sparse singular value decomposition . - 97 p.

Source: Dissertation Abstracts International, Volume: 71-03, Section: B, page: 1821.

Thesis (Ph.D.)--University of Florida, 2009.

Singular value decomposition is a problem that is used in a wide variety of applications like latent semantic indexing, collaborative filtering and gene expression analysis. In this study, we consider the singular value decomposition problem for band and sparse matrices. Linear algebraic algorithms for modern computer architectures are designed to extract maximum performance by exploiting modern memory hierarchies, even though this can sometimes lead to algorithms with higher memory requirements and more floating point operations. We propose blocked algorithms for sparse and band bidiagonal reduction.

ISBN: 9781109670950Subjects--Topical Terms:

1669109
Applied Mathematics.

Efficient algorithms for sparse singular value decomposition .
LDR:03059nam 2200313 4500 001 1391033
005 20101222085254.5
008 130515s2009 ||||||||||||||||| ||eng d
020 $a 9781109670950
035 $a (UMI)AAI3400303
035 $a AAI3400303
040 $a UMI $c UMI
100 1 $a Rajamanickam, Sivasankaran. $3 1669405
245 1 0 $a Efficient algorithms for sparse singular value decomposition .
300 $a 97 p.
500 $a Source: Dissertation Abstracts International, Volume: 71-03, Section: B, page: 1821.
500 $a Adviser: Timothy A. Davis.
502 $a Thesis (Ph.D.)--University of Florida, 2009.
520 $a Singular value decomposition is a problem that is used in a wide variety of applications like latent semantic indexing, collaborative filtering and gene expression analysis. In this study, we consider the singular value decomposition problem for band and sparse matrices. Linear algebraic algorithms for modern computer architectures are designed to extract maximum performance by exploiting modern memory hierarchies, even though this can sometimes lead to algorithms with higher memory requirements and more floating point operations. We propose blocked algorithms for sparse and band bidiagonal reduction.
520 $a The blocked algorithms are designed to exploit the memory hierarchy, but they perform nearly the same number of floating point operations as the non-blocked algorithms. We introduce efficient blocked band reduction algorithms that utilize the cache correctly and perform better than competing methods in terms of the number of floating point operations and the amount of required workspace. Our band reduction methods are several times faster than existing methods.
520 $a The theory and algorithms for sparse singular value decomposition, especially algorithms for reducing a sparse upper triangular matrix to a bidiagonal matrix are proposed here. The bidiagonal reduction algorithms use a dynamic blocking method to reduce more than one entry at a time. They limit the sub-diagonal fill to one scalar by pipelining the blocked plane rotations. A symbolic factorization algorithm for computing the time and memory requirements for the bidiagonal reduction of a sparse matrix helps the numerical reduction step.
520 $a Our sparse singular value decomposition algorithm computes all the singular values at the same amount of time it takes to compute a few singular values using existing methods. It performs much faster than existing methods when more singular values are required. The features of the software implementing the band and sparse bidiagonal reduction algorithms are also presented. (Full text of this dissertation may be available via the University of Florida Libraries web site. Please check http://www.uflib.ufl.edu/etd.html)
590 $a School code: 0070.
650 4 $a Applied Mathematics. $3 1669109
650 4 $a Computer Science. $3 626642
690 $a 0364
690 $a 0984
710 2 $a University of Florida. $3 718949
773 0 $t Dissertation Abstracts International $g 71-03B.
790 1 0 $a Davis, Timothy A., $e advisor
790 $a 0070
791 $a Ph.D.
792 $a 2009
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3400303