Efficient Randomized Algorithms for the Fixed Precision Low Rank Matrix Approximation
The Author may post the final published version of the Work on the Author's personal web site and on the web server of the Author's institution, provided that proper notice of the Publisher's copyright is included and that no separate or additional fees are collected for access to or distribution of the work.
© 2018, Society for Industrial and Applied Mathematics (SIAM)
Unauthorized reproduction of this article is prohibited.
Randomized algorithms for low-rank matrix approximation are investigated, with the emphasis on the fixed-precision problem and computational efficiency for handling large matrices. The algorithms are based on the so-called QB factorization, where Q is an orthonormal matrix. First, a mechanism for calculating the approximation error in the Frobenius norm is proposed, which enables efficient adaptive rank determination for a large and/or sparse matrix. It can be combined with any QB-form factorization algorithm in which B's rows are incrementally generated. Based on the blocked randQB algorithm by Martinsson and Voronin, this results in an algorithm called randQB_EI. Then, we further revise the algorithm to obtain a pass-efficient algorithm, randQB_FP, which is mathematically equivalent to the existing randQB algorithms and also suitable for the fixed-precision problem. Especially, randQB_FP can serve as a single-pass algorithm for calculating leading singular values, under a certain condition. With large and/or sparse test matrices, we have empirically validated the merits of the proposed techniques, which exhibit remarkable speedup and memory saving over the blocked randQB algorithm. We have also demonstrated that the single-pass algorithm derived by randQB_FP is much more accurate than an existing single-pass algorithm. And with data from a scenic image and an information retrieval application, we have shown the advantages of the proposed algorithms over the adaptive range finder algorithm for solving the fixed-precision problem.