This question was automatically deleted from
math.stackexchange.com.
Question
Matrices estimated from data, even when, are not always invertible. Fortunately a ‘nearest inverse’ can be calculated with the Moore-Penrose inverse. Recently I have been taking inverses of tensors of rank greater than two that are calculated from real-world data, but I am finding that they are often non-invertible.
Is there a pseudoinverse for finitary rank tensors in a similar sense to the Moore-Penrose pseudoinverse?