Digital image processing
A Parallel Algorithm for Computing the Singular Value Decomposition of a Matrix
SIAM Journal on Matrix Analysis and Applications
A Stable and Efficient Algorithm for the Rank-One Modification of the Symmetric Eigenproblem
SIAM Journal on Matrix Analysis and Applications
A Divide-and-Conquer Algorithm for the Bidiagonal SVD
SIAM Journal on Matrix Analysis and Applications
A Divide-and-Conquer Algorithm for the Symmetric TridiagonalEigenproblem
SIAM Journal on Matrix Analysis and Applications
The instability of parallel prefix matrix multiplication
SIAM Journal on Scientific Computing
Parallel programming with MPI
Applied numerical linear algebra
Applied numerical linear algebra
ScaLAPACK user's guide
On Computing an Eigenvector of a Tridiagonal Matrix. Part I: Basic Results
SIAM Journal on Matrix Analysis and Applications
LAPACK Users' guide (third ed.)
LAPACK Users' guide (third ed.)
Parallel Complexity of Linear System Solution
Parallel Complexity of Linear System Solution
Orthogonal Eigenvectors and Relative Gaps
SIAM Journal on Matrix Analysis and Applications
Parallelism of double divide and conquer algorithm for singular value decomposition
PDCN'07 Proceedings of the 25th conference on Proceedings of the 25th IASTED International Multi-Conference: parallel and distributed computing and networks
A new algorithm for singular value decomposition and its parallelization
Parallel Computing
Parallel double divide and conquer and its evaluation on a super computer
PDCS '07 Proceedings of the 19th IASTED International Conference on Parallel and Distributed Computing and Systems
Hi-index | 0.00 |
This paper proposes a new singular value decomposition algorithm which can be fully parallelized. I-SVD algorithm was recently proposed using the theory of integrable systems. It computes singular value decompositions much faster than standard algorithms such as QR and Divide and Conquer(D&C), and at the same accuracy level. However, we show that the parallelism of I-SVD is limited to singular vector computation only because it shows seriality in the case of singular value computation. Our concern is a fully parallelizable algorithm whose serial version is as fast and accurate as I-SVD. Compared to the original D&C, where most of the running time is consumed for vector updating during singular vector computation, our new algorithm first computes singular values by the "compact" D&C without computing singular vectors. Thus, the compact D&C is wholly faster than the original one. The corresponding singular vectors are then computed by twisted factorization. The algorithm has great parallelism because each step can be executed parallelly. Our new algorithm is numerically tested on some SVD computations. It is as fast as I-SVD and even much faster than the standard algorithms. Also, its accuracy is as good as that of the others.