“…Various authors demonstrated the capacity to compensate for the nonlinear distortion of these methods, but their high computational complexity prevented them from being adopted in real-time applications. In order to find a trade-off between performance and computational complexity, different machine learning methods have been proposed, including clustering [11], [12], [13], [14], supervised regression [15], [16], [17], [18], and supervised classification [19], [20]. Supervised classification is particularly interesting because, on the one hand, it typically leads to better performance than clustering.…”