The higher-order power method revisited: convergence proofs and effective initialization
暂无分享,去创建一个
We revisit the higher-order power method of De Lathauwer et al. (1995) for rank-one tensor approximation, and its relation to contrast maximization as used in blind deconvolution. We establish a simple convergence proof for the general nonsymmetric tensor case. We show also that a symmetric version of the algorithm, offering an order of magnitude reduction in computational complexity but discarded by De Lathauwer et al. as unpredictable, is likewise provably convergent. A new initialization scheme is also developed which, unlike the TSVD-based initialization, leads to a quantifiable proximity to the globally optimal solution.
[1] S. R. Searle,et al. The Vec-Permutation Matrix, the Vec Operator and Kronecker Products: A Review , 1981 .
[2] Ehud Weinstein,et al. New criteria for blind deconvolution of nonminimum phase systems (channels) , 1990, IEEE Trans. Inf. Theory.
[3] P. Regalia,et al. Tensor displacement structures and polyspectral matching , 1999 .
[4] Phillip A. Regalia,et al. Undermodeled equalization in noisy multi-user channels , 1999, NSIP.