A BLOCK BIDIAGONALIZATION METHOD FOR FIXED-ACCURACY LOW-RANK MATRIX APPROXIMATION

We present randUBV, a randomized algorithm for matrix sketching based on the block Lanzcos bidiagonalization process. Given a matrix A, it produces a low-rank approximation of the form UBVT, where U and V have orthonormal columns in exact arithmetic and B is block bidiagonal. In finite precision, the columns of both U and V will be close to orthonormal. Our algorithm is closely related to the randQB algorithms of Yu, Gu, and Li [SIAM J. Matrix Anal. Appl., 39 (2018), pp. 1339-1359]. in that the entries of B are incrementally generated and the Frobenius norm approximation error may be efficiently estimated. It is therefore suitable for the fixed-accuracy problem and so is designed to terminate as soon as a user input error tolerance is reached. Numerical experiments suggest that the block Lanczos method is generally competitive with or superior to algorithms that use power iteration, even when A has significant clusters of singular values.