un 2 00 9 Subtracting a best rank - 1 approximation may increase tensor rank
نویسندگان
چکیده
It has been shown that a best rank-R approximation of an order-k tensor may not exist when R ≥ 2 and k ≥ 3. This poses a serious problem to data analysts using tensor decompositions. It has been observed numerically that, generally, this issue cannot be solved by consecutively computing and subtracting best rank-1 approximations. The reason for this is that subtracting a best rank-1 approximation generally does not decrease tensor rank. In this paper, we provide a mathematical treatment of this property for real-valued 2 × 2 × 2 tensors, with symmetric tensors as a special case. Regardless of the symmetry, we show that for generic 2 × 2 × 2 tensors (which have rank 2 or 3), subtracting a best rank-1 approximation results in a tensor that has rank 3 and lies on the boundary between the rank-2 and rank-3 sets. Hence, for a typical tensor of rank 2, subtracting a best rank-1 approximation increases the tensor rank. 1 A 5-page summary of this paper has been accepted for the EUSIPCO 2009 conference and will appear in its proceedings.
منابع مشابه
Subtracting a best rank-1 approximation does not necessarily decrease tensor rank
It has been shown that a best rank-R approximation of an order-k tensor may not exist when R ≥ 2 and k ≥ 3. This poses a serious problem to data analysts using tensor decompositions. It has been observed numerically that, generally, this issue cannot be solved by consecutively computing and subtracting best rank-1 approximations. The reason for this is that subtracting a best rank-1 approximati...
متن کاملBest Rank-One Tensor Approximation and Parallel Update Algorithm for CPD
A novel algorithm is proposed for CANDECOMP/PARAFAC tensor decomposition to exploit best rank-1 tensor approximation. Different from the existing algorithms, our algorithm updates rank-1 tensors simultaneously in-parallel. In order to achieve this, we develop new all-at-once algorithms for best rank-1 tensor approximation based on the Levenberg-Marquardt method and the rotational update. We sho...
متن کاملComplex Tensors Almost Always Have Best Low-rank Approximations
Low-rank tensor approximations are plagued by a well-known problem — a tensor may fail to have a best rank-r approximation. Over R, it is known that such failures can occur with positive probability, sometimes with certainty: in R2×2×2, every tensor of rank 3 fails to have a best rank-2 approximation. We will show that while such failures still occur over C, they happen with zero probability. I...
متن کاملWell-posedness of convex maximization problems on Stiefel manifolds and orthogonal tensor product approximations
Problems of best tensor product approximation of low orthogonal rank can be formulated as maximization problems on Stiefel manifolds. The functionals that appear are convex and weakly sequentially continuous. It is shown that such problems are always well-posed, even in the case of non-compact Stiefel manifolds. As a consequence, problems of finding a best orthogonal, strong orthogonal or compl...
متن کاملRelative Error Tensor Low Rank Approximation
We consider relative error low rank approximation of tensors with respect to the Frobenius norm. Namely, given an order-q tensor A ∈ R ∏q i=1 ni , output a rank-k tensor B for which ‖A − B‖F ≤ (1 + ) OPT, where OPT = infrank-k A′ ‖A − A‖F . Despite much success on obtaining relative error low rank approximations for matrices, no such results were known for tensors. One structural issue is that ...
متن کامل