You're comparing several supervised learning models with different levels of complexity and want to choose one that will perform well on unseen data.
How would you explain the trade-offs between model complexity and generalization?
