Abstract:A random submatrix method (RSM) is proposed to calculate the low-rank decomposition Ûmxr V̂nxrT (r <; m, n) of the matrix Y ∈ Rmxn (assuming m > n generally) with known entry percentage 0 <; p ≤ 1. RSM is very fast as only O(mr2pr) or O(n3p3r) floating-point operations (flops) are required, compared favorably with O(mnr + r2(m + n)) flops required by the state-of-the-art algorithms. Meanwhile, RSM has the advantage of a small memory requirement as only max(n2, mr + nr) real values need to be saved. With the assumption that known entries are uniformly distributed in Y, submatrices formed by known entries are randomly selected from Y with statistical size k x npk or mpl x l, where k or l takes r + 1 usually. We propose and prove a theorem, under random noises the probability that the subspace associated with a smaller singular value will turn into the space associated to anyone of the r largest singular values is smaller. Based on the theorem, the npk - k null vectors or the l - r right singular vectors associated with the minor singular values are calculated for each submatrix. The vectors ought to be the null vectors of the submatrix formed by the chosen npk or l columns of the ground truth of V̂T. If enough submatrices are randomly chosen, V̂ and Û can be estimated accordingly. The experimental results on random synthetic matrices with sizes such as 131072 x 1024 and on real data sets such as dinosaur indicate that RSM is 4.30 ~ 197.95 times faster than the state-of-the-art algorithms. It, meanwhile, has considerable high precision achieving or approximating to the best.

Structure-Aware Analyses and Algorithms for Interpolative Decompositions

A QLP Decomposition via Randomization.

A general error analysis for randomized low-rank approximation with application to data assimilation

Robust Blockwise Random Pivoting: Fast and Accurate Adaptive Interpolative Decomposition

A Quadratically Convergent Algorithm for Structured Low-Rank Approximation

Randomized algorithms for low-rank matrix approximation: Design, analysis, and applications

Adaptive Parallelizable Algorithms for Interpolative Decompositions via Partially Pivoted LU

Randomized Rank-Revealing QLP for Low-Rank Matrix Decomposition

Learning-Based Low-Rank Approximations

Single-pass randomized QLP decomposition for low-rank approximation

A Random Algorithm for Low-Rank Decomposition of Large-Scale Matrices with Missing Entries

On the Unreasonable Effectiveness of Single Vector Krylov Methods for Low-Rank Approximation

Towards Efficient and Accurate Approximation: Tensor Decomposition Based on Randomized Block Krylov Iteration

On the analysis of optimization with fixed-rank matrices: a quotient geometric view

Towards a more robust algorithm for computing the restricted singular value decomposition

A Randomized Algorithm for Approximating Truncated SVD

Randomized GCUR decompositions

Regression-aware decompositions

Fast and Accurate Randomized Algorithms for Low-rank Tensor Decompositions

The restricted singular value decomposition: properties and applications