NANAOct 7, 2018

Interpolative Decomposition Butterfly Factorization

arXiv:1809.1057314 citations
AI Analysis

For researchers and practitioners needing fast matrix-vector multiplication for large matrices with complementary low-rank structure, this provides a general, kernel-independent framework with near-optimal complexity.

This paper introduces a kernel-independent interpolative decomposition butterfly factorization (IDBF) for matrices with complementary low-rank property, achieving O(N log N) construction and application time. Numerical results demonstrate its effectiveness for fast matrix-vector multiplication in applications like Fourier integral operators and high-frequency wave computation.

This paper introduces a "kernel-independent" interpolative decomposition butterfly factorization (IDBF) as a data-sparse approximation for matrices that satisfy a complementary low-rank property. The IDBF can be constructed in $O(N\log N)$ operations for an $N\times N$ matrix via hierarchical interpolative decompositions (IDs), if matrix entries can be sampled individually and each sample takes $O(1)$ operations. The resulting factorization is a product of $O(\log N)$ sparse matrices, each with $O(N)$ non-zero entries. Hence, it can be applied to a vector rapidly in $O(N\log N)$ operations. IDBF is a general framework for nearly optimal fast matvec useful in a wide range of applications, e.g., special function transformation, Fourier integral operators, high-frequency wave computation. Numerical results are provided to demonstrate the effectiveness of the butterfly factorization and its construction algorithms.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes