You're comparing several candidate models for a supervised learning task and want to improve performance without overfitting.
How would you optimize a machine learning model for performance?
