The optimization of mix-valued probabilistic logical networks is a natural extension of optimization of Boolean networks. In this study we have first obtained a recursive solution for the finite horizon case. Then we have proved that when the filter length is large enough, the obtained optimal control sequence coincides with the one for the infinite horizon case using the reeding horizon technique. This result turns searching an infinite sequence of controls into finding an optimal feedback matrix by solving a finite horizon optimization problem. As examples, its applications to human-machine game and to metastatic melanoma are investigated.