Efficient coordinate-wise leading eigenvector computation
Efficient coordinate-wise leading eigenvector computation
复制标题
DOI:
--
复制
发表时间:
2017-02
期刊:
影响因子:
--
通讯作者:
Jialei Wang;Weiran Wang;D. Garber;N. Srebro
中科院分区:
文献类型:
--
作者:
Jialei Wang;Weiran Wang;D. Garber;N. Srebro
We develop and analyze efficient "coordinate-wise" methods for finding the leading eigenvector, where each step involves only a vector-vector product. We establish global convergence with overall runtime guarantees that are at least as good as Lanczos's method and dominate it for slowly decaying spectrum. Our methods are based on combining a shift-and-invert approach with coordinate-wise algorithms for linear regression.