You're comparing several supervised learning models and want a principled way to choose the right level of model complexity.
What is the bias-variance tradeoff, and how does it affect model selection?
