On the Largest Multilinear Singular Values of Higher-Order Tensors.

SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS(2017)

引用 7|浏览5
暂无评分
摘要
Let an denote the largest mode-n multilinear singular value of an I-1 x ... x I-N tensor tau. We prove that sigma(2)(1) + ... + sigma(2)(n-1) +sigma(2)(n+1) + ... + sigma(2)(N) <= (N- 2) parallel to tau parallel to(2) + sigma(2)(n), n = 1, ... , N, where parallel to.parallel to denotes the Frobenius norm. We also show that at least for third-order cubic tensors the inverse problem always has a solution. Namely, for each sigma(1), sigma(2), and sigma(3) that satisfy sigma(2)(1) + sigma(2)(2) <= parallel to tau parallel to(2) + sigma(2)(3), sigma(2)(1) + sigma(2)(3) <= parallel to tau parallel to(2) + sigma(2)(2), sigma(2)(2) + sigma(2)(3) <= parallel to tau parallel to(2) + sigma(2)(1), and the trivial inequalities sigma(1) >= 1/root n parallel to tau parallel to, sigma(2) >= 1/root n parallel to tau parallel to, sigma(3) >= 1/root n parallel to tau parallel to, there always exists an n x n x n tensor whose largest multilinear singular values are equal to sigma(1), sigma(2), and sigma(3). We also show that if the equality sigma(2)(1) + sigma(2)(2) = parallel to tau parallel to(2) + sigma(2)(3) holds, then tau is necessarily equal to a sum of multilinear rank-(L-1, 1, L-1) and multilinear rank-(1, L-2, L-2) tensors and we give a complete description of all its multilinear singular values. We establish a connection with honeycombs and eigenvalues of the sum of two Hermitian matrices. This seems to give at least a partial explanation of why results on the joint distribution of multilinear singular values are scarce.
更多
查看译文
关键词
multilinear singular value decomposition,multilinear rank,singular value decomposition,tensor
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要