A New Paradigm of Precoder Design for Finite-Alphabet Inputs Based Communications

This paper proposes a pair of new linear precoder designs to reduce the high computation complexity in the context of multi-input multi-output communications with finite-alphabet inputs. We initially develop a high-performance mutual information maximization scheme using closed cut-off rate expression, formulating a standard convex optimization problem for iterative precoder optimization. To address the computational challenges associated with a large number of the possible transmit symbol vectors, we leverage the convexity of the exponential function to simplify the objective function. This approach results in an efficient bound that depends only on the maximum and minimum terms. Consequently, the corresponding optimization problem requires only two auxiliary variables to optimize the precoder, significantly reducing the computational complexity. Finally, numerical simulation verifies the efficiency of our proposed mutual information maximization schemes in terms of performance-versus-computation tradeoff.