Accelerated alternating MIMO optimization

Various majorization approaches allow to transform a Weighted Sum Rate criterion into a cost function that is quadratic in Tx or in Rx and is simple to optimize in terms of other parameters. The standard strategy is then to apply alternating minimization to the majorizer. These alternating approaches converge fairly slowly. Another issue that they typically are guaranteed to converge, but to a local optimum. Deterministic annealing has been proposed to find the global optimum. To reduce convergence time, acceleration techniques can be introduced. In this paper we review existing acceleration methods. Two big families arise, either Nesterov acceleration or Successive Over-Relaxation. The question arises and remains open which approach would be best suited to alternating precoder design.