which avoids the use of the inverse of W. Then Yuan borrowed the idea of splitting the form of A and W
from [22], and took some special block preconditioned matrices to get a couple of reduced systems which
can be easily solved by block SOR methods and the conjugate gradient method. Specially the conjugate
gradient method to solve problem (2.1) by a long matrix–vector product instead of inverse matrix W 1.
Then he studied convergence theory for such methods for solving problem (2.1) [38]. The convergence results are generalization of convergence results of such methods for solving the least squares problems
where W ¼ I given in [22] because Yuan’s results in [38,40,40,39] reduce to the results for least squares
problems given in [22].