Effect of Sampling Method on the Regression Accuracy for a High-Speed Link Problem

We examine the effect of different sampling methods on the accuracy of support vector regression models for eye opening prediction of a high-speed link. Four different sampling methods are tested to generate training data for training the machine learning model for the equalizers on a high-speed link problem, and the accuracy of the models to predict eye opening is compared. Latin Hypercube shows superior performance in terms of mean square error and
$R^{2}$
compared to other methods tested.