We often use non-linear and non-parametric models in order to build robust models, and yet the majority of people choose to use PCA plots for visualization. t-SNE is a great option for data sets which are a little more complicated, an a great R and Python tutorial can be found here.