The Block Parallel Computation of Matrix Tensor Production

摘要

In matrix computation, the computation of matrix tensor product is an important problem. Compared with its multiplication, the computing amount of tensor product is huger. Based on analyzing its mathematical properties, a result is proved that two matrices tensor operation can be exchanged in the sense of permuted similar, and the convenient method for constructing the permutation matrix is given. Furthermore, a conclusion is obtained that tensor product of block matrixes can be block calculated in the sense of permuted similar. Based on these, the parallel computing models of matrix tensor product are proposed. From an example, the thought and process of the algorithm are showed.