Discrete optimization for error feedback network using lower bound estimation

We propose a new design method for error feedback (EF) with discrete coefficients. The lower bound estimation principle is used in order to search for the optimum solution and the EF networks designed by the proposed method have guaranteed optimality. Additionally, the method can reduce the calculation cost because we can cut the sub-trees based on the lower bound estimation. Finally, we present a numerical example by designing the (3, 3)th order EF with discrete coefficients whose decimal parts are 2 bits. From the results, we demonstrate the effectiveness of the proposed method and show that the calculation time is a few seconds.