Japan Journal of Industrial and Applied Mathematics ( IF 0.7 ) Pub Date : 2021-02-24 , DOI: 10.1007/s13160-021-00459-x Haruka Kawamura , Reiji Suda
Low-rank approximation by QR decomposition with pivoting (pivoted QR) is known to be less accurate than singular value decomposition (SVD); however, the calculation amount is smaller than that of SVD. The least upper bound of the ratio of the truncation error, defined by \(\Vert A-BC\Vert _2\), using pivoted QR to that using SVD is proved to be \(\sqrt{\frac{4^k-1}{3}(n-k)+1}\) for \(A\in {\mathbb {R}}^{m\times n}\) \((m\ge n)\), approximated as a product of \(B\in {\mathbb {R}}^{m\times k}\) and \(C\in {\mathbb {R}}^{k\times n}\) in this study.
中文翻译:
基于枢轴QR分解的低秩矩阵逼近算法的截断误差的最小上限
已知通过旋转旋转QR分解(旋转QR)进行的低秩逼近比奇异值分解(SVD)精度低;但是,计算量小于SVD。证明由\(\ Vert A-BC \ Vert _2 \)定义的使用旋转QR的截断误差与使用SVD的截断误差的比率的最小上限为\(\ sqrt {\ frac {4 ^ k- 1} {3}(NK)+1} \)为\(A \在{\ mathbb {R}} ^ {米\ n次} \) \((M \ GE N)\),近似为产品本研究中\(B \ in {\ mathbb {R}} ^ {m \ times k} \)和\(C \ in {\ mathbb {R}} ^ {k \ times n} \)中的分布。