Conjugate Gradient Algorithm for the Symmetric Arrowhead Solution of Matrix Equation AXB=C

2 A denote the transpose, Moore-Penrose generalized inverse, Frobenius norm and Euclid norm, respectively. For any , m n A B  R ,   , 0 T A B trace B A   denotes the inner product of A and B . Therefore, m n R  is a complete inner product space endowed with 2 , A A A  . For any non-zero matrices 1 2 , , , m n k A A A  R  , if  , T j i i A A trace A     0 j A i j   , then it is easy to verify that 1 2 , , , k A A A  are linearly independent and orthogonal. Proposition 1. Let , n n A B  R , then ( ) ( ); T trace A trace A  ( ) ( ) trace AB trace BA  ( ) ( ) ( ) trace A B trace A trace B   