On the extension of trace norm to tensors

In this paper, we propose three extensions of trace norm for the minimization of tensor rank via convex optimization. The alternating direction method of multipliers is used to efficiently solve the optimization problems. One of the proposed extensions recovers partially observed tensor almost perfectly from a small fraction of observations.