Gaussian Processes and its Application to the design of Digital Communication Receivers
Pablo M., Juan José, Fernando Pérez‐Cruz · InTech eBooks · 2010
We have proposed GPR and GPC for designing digital communication receivers. GPR follows a wide range of machine learning tools that have been successfully applied to the design of digital communication receivers. GPR can be viewed as a nonlinear MMSE. MMSE is the standard criterion used for designing digital communication receivers, as it trades off inverting the channel and not amplifying the noise. GPR solution is analytical given the nonlinear function, while most machine-learning methods need to perform an optimization problem to achieve their solution. On the other hand, GPC provides extra information for each one of its decisions, i.e. the posterior probability of being in the correct class. This information can be used by the channel decoder to significantly reduce the BER for low signal to noise ratio. This characteristic is not shared by the other nonlinear machine learning tools, as they can only provide hard decisions as outputs. We have shown that, as the number of samples increases, the predicted probabilities tend to the true posterior probabilities. To highlight the advantages of GPs as digital communications receivers we compare their performances to that of SVM. SVM provides solutions as good as GPR does, but it needs more training samples. These tools have been compared in two typical scenarios in digital communications: equalization and multiuser detection. In both experiments GPs exhibit an outstanding behavior.