Multiple-Rank Updates to Matrix Factorizations
Zack 8/30/2013
Multiple-Rank Updates to Matrix Factorizations Zack 8/30/2013 - - PowerPoint PPT Presentation
Multiple-Rank Updates to Matrix Factorizations Zack 8/30/2013 Outline u Introduction u Multiple-rank Update Theory u Sparse Matrix Factorization by Multiple-rank Update u Symmetric Factorization by Multiple-rank Update u Conclusion Introduction
Zack 8/30/2013
u Introduction u Multiple-rank Update Theory u Sparse Matrix Factorization by Multiple-rank Update u Symmetric Factorization by Multiple-rank Update u Conclusion
u Dirext method
u Left-looking method u Right-looking method
u Computation effort
u O(n3) for dense matrix u O(n1.05-2) for sparse matrix
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
u
Methods Flops Comments
Left-looking LU O(n2) memory write, bad in parallel Right-looking LU O(n3) memory write, good in parallel QR O(n3) memory write, good in parallel n rank-1 updates O(n3) memory write, good in parallel
u
u
u The path of a rank-1 update follow exactly the path of the
elimination tree.
u Tridiagonal matrix’s elimination tree.
u Base-camp Theory
u
u
u
u
u
u
u
u
u
u
u Introduced multiple 1-rank update method u Optimized the method when matrix is sparse. u Optimized the method when matrix is symmetry.
u Deng L. Multiple-rank Updates to Matrix Factorizations for Nonlinear Analysis
and Circuit Design[D]. Stanford University, 2010.