We give explicit low-rank bilinear non-commutative schemes for multiplying structured $n \times n$ matrices with $2 \leq n \leq 5$, which serve as building blocks for recursive algorithms with improved multiplicative factors in asymptotic complexity. Our schemes are discovered over $\mathbb{F}_2$ or $\mathbb{F}_3$ and lifted to $\mathbb{Z}$ or $\mathbb{Q}$. Using a flip graph search over tensor decompositions, we derive schemes for general, upper-triangular, lower-triangular, symmetric, and skew-symmetric inputs, as well as products of a structured matrix with its transpose. These schemes improve asymptotic constants for 13 of 15 structured formats. In particular, we obtain $4 \times 4$ rank-34 schemes for both multiplying a general matrix by its transpose and an upper-triangular matrix by a general matrix, improving the asymptotic factor from 8/13 (0.615) to 22/37 (0.595). Additionally, using $\mathbb{F}_3$ flip graphs, we discover schemes over $\mathbb{Q}$ that fundamentally require the inverse of 2, including a $2 \times 2$ symmetric-symmetric multiplication of rank 5 and a $3 \times 3$ skew-symmetric-general multiplication of rank 14 (improving upon AlphaTensor's 15).
翻译:我们针对维度为 $2 \leq n \leq 5$ 的结构化 $n \times n$ 矩阵乘法,提出了显式的低秩双线性非交换计算方案,这些方案可作为递归算法的构建模块,以改进渐近复杂度中的乘法因子。我们的方案在 $\mathbb{F}_2$ 或 $\mathbb{F}_3$ 上被发现,并提升至 $\mathbb{Z}$ 或 $\mathbb{Q}$。通过对张量分解进行翻转图搜索,我们推导出适用于一般矩阵、上三角矩阵、下三角矩阵、对称矩阵及反对称矩阵输入的计算方案,以及结构化矩阵与其转置乘积的方案。这些方案在15种结构化格式中的13种上改进了渐近常数。特别地,我们获得了 $4 \times 4$ 秩-34方案,用于一般矩阵与其转置的乘法以及上三角矩阵与一般矩阵的乘法,将渐近因子从8/13(0.615)提升至22/37(0.595)。此外,利用 $\mathbb{F}_3$ 翻转图,我们发现了在 $\mathbb{Q}$ 上本质上需要2的逆元的方案,包括秩为5的 $2 \times 2$ 对称-对称乘法和秩为14的 $3 \times 3$ 反对称-一般乘法(优于AlphaTensor的15秩方案)。