Classification

Feature selection is just as important as hyperparameter choice, yet many classifiers cherry-pick features without much empirical basis. While hyperparameters are easier to tweak, we must account for the impact of boosting these hyperparameters and the risk and impact of overfit.