Pass-efficient truncated UTV for low-rank approximations

We propose the pass-efficient truncated UTV algorithm, a faster variant of the TUXV algorithm for low-rank approximations. Compared with the TUXV algorithm, data transfer and the complexity of the proposed algorithm are reduced. Therefore, our algorithm is suitable for large matrices stored out of memory or generated by streaming data. We also develop residual error upper bounds and singular value approximation error bounds for the pass-efficient truncated UTV algorithm. Numerical experiments are reported to demonstrate the efficiency and effectiveness of our algorithm.