Success Probabilities of L2-norm Regularized Babai Detectors and Maximization

This paper is concerned with success probabilities of some L
2
-norm regularized Babai detectors for MIMO detection. First, we extend the often used MMSE-SIC detector to two L
2
-norm regularized Babai detectors. Then, we derive the corresponding success probability formulae of the two detectors. To increase their success probabilities, we propose to maximize them with respect to the regularization parameter vector by a machine learning based method called ML-λ, as conventional optimization methods are very inefficient. This moves the exponential computational cost to offline training, and the method outputs an approximation to the optimal regularization parameter vector in polynomial time. Finally, we present some numerical simulations to demonstrate the performance of the new detectors.