On the average condition number of tensor rank decompositions
arXiv:1801.01673 · doi:10.1093/imanum/drz026
Abstract
We compute the expected value of powers of the geometric condition number of random tensor rank decompositions. It is shown in particular that the expected value of the condition number of tensors with a random rank- decomposition, given by factor matrices with independent and identically distributed standard normal entries, is infinite. This entails that it is expected and probable that such a rank- decomposition is sensitive to perturbations of the tensor. Moreover, it provides concrete further evidence that tensor decomposition can be a challenging problem, also from the numerical point of view. On the other hand, we provide strong theoretical and empirical evidence that tensors of size with all have a finite average condition number. This suggests there exists a gap in the expected sensitivity of tensors between those of format and other order-3 tensors. For establishing these results, we show that a natural weighted distance from a tensor rank decomposition to the locus of ill-posed decompositions with an infinite geometric condition number is bounded from below by the inverse of this condition number. That is, we prove one inequality towards a so-called condition number theorem for the tensor rank decomposition.
References in corpus (10)
- Tensor Decomposition for Signal Processing and Machine Learning
- Identifiability of parameters in latent structure models with many observed variables
- An algorithm for generic and low-rank specific identifiability of complex tensors
- Generic uniqueness conditions for the canonical polyadic decomposition and INDSCAL
- Effective criteria for specific identifiability of tensors and forms
- The condition number of join decompositions
- A Riemannian Trust Region Method for the Canonical Tensor Rank Approximation Problem
- A condition number for the tensor rank decomposition
- Pencil-based algorithms for tensor rank decomposition are not stable
- Convergence analysis of Riemannian Gauss-Newton methods and its connection with the geometric condition number