Tensor completion by multi-rank via unitary transformation

One of the key problems in tensor completion is number uniformly random sample entries required for recovery guarantee. The main aim this paper to study n1×n2×n3 third-order based on transformed singular value decomposition, and provide a bound entries. Our approach make use multi-rank underlying instead its tubal rank bound. In numerical experiments synthetic imaging data sets, we demonstrate effectiveness our proposed Moreover, theoretical results are valid any unitary transformation applied n3-dimension under decomposition.