Machine-Learning-Based Optimization of Tx Equalization Parameters of a High-Speed Channel

We propose a machine-learning-based optimization approach for the Tx equalization. Our target is practical high-speed channels used in systems requiring industrial communication protocols. We adopted Random Forest as our basic machine learning algorithm and applied it to low, medium, and high loss channels with 3-tap and 5tap Tx equalization architectures. Our machine-learning-based optimization results showed the outstanding efficiency and accuracy of the approach.