Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2020
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2003.03019 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915602810011648 |
|---|---|
| author | Christandl, Matthias Gall, François Le Lysikov, Vladimir Zuiddam, Jeroen |
| author_facet | Christandl, Matthias Gall, François Le Lysikov, Vladimir Zuiddam, Jeroen |
| contents | We study the algorithmic problem of multiplying large matrices that are rectangular. We prove that the method that has been used to construct the fastest algorithms for rectangular matrix multiplication cannot give algorithms with complexity $n^{p + 1}$ for $n \times n$ by $n \times n^p$ matrix multiplication. In fact, we prove a precise numerical barrier for this method. Our barrier improves the previously known barriers, both in the numerical sense, as well as in its generality. In particular, we prove that any lower bound on the dual exponent of matrix multiplication $α$ via the big Coppersmith-Winograd tensors cannot exceed 0.6218. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2003_03019 |
| institution | arXiv |
| publishDate | 2020 |
| record_format | arxiv |
| spellingShingle | Barriers for rectangular matrix multiplication Christandl, Matthias Gall, François Le Lysikov, Vladimir Zuiddam, Jeroen Computational Complexity Commutative Algebra 68Q17, 15A69 We study the algorithmic problem of multiplying large matrices that are rectangular. We prove that the method that has been used to construct the fastest algorithms for rectangular matrix multiplication cannot give algorithms with complexity $n^{p + 1}$ for $n \times n$ by $n \times n^p$ matrix multiplication. In fact, we prove a precise numerical barrier for this method. Our barrier improves the previously known barriers, both in the numerical sense, as well as in its generality. In particular, we prove that any lower bound on the dual exponent of matrix multiplication $α$ via the big Coppersmith-Winograd tensors cannot exceed 0.6218. |
| title | Barriers for rectangular matrix multiplication |
| topic | Computational Complexity Commutative Algebra 68Q17, 15A69 |
| url | https://arxiv.org/abs/2003.03019 |