Tensor Completion Made Practical
This addresses the impracticality of existing tensor completion methods for real-world applications, offering a provably efficient and scalable solution.
The paper tackles the problem of tensor completion, which recovers low-rank tensors from sparse observations, by introducing a new variant of alternating minimization that converges linearly even with highly correlated factors and is practical for large-scale tensors, such as completing third-order tensors with a thousand dimensions from a tiny fraction of entries.
Tensor completion is a natural higher-order generalization of matrix completion where the goal is to recover a low-rank tensor from sparse observations of its entries. Existing algorithms are either heuristic without provable guarantees, based on solving large semidefinite programs which are impractical to run, or make strong assumptions such as requiring the factors to be nearly orthogonal. In this paper we introduce a new variant of alternating minimization, which in turn is inspired by understanding how the progress measures that guide convergence of alternating minimization in the matrix setting need to be adapted to the tensor setting. We show strong provable guarantees, including showing that our algorithm converges linearly to the true tensors even when the factors are highly correlated and can be implemented in nearly linear time. Moreover our algorithm is also highly practical and we show that we can complete third order tensors with a thousand dimensions from observing a tiny fraction of its entries. In contrast, and somewhat surprisingly, we show that the standard version of alternating minimization, without our new twist, can converge at a drastically slower rate in practice.