There are 3 techniques to control the complexity of a machine learning model. The first is to dial a complexity hyperparameter which does not allow you to choose a very complex model. The second is to carry our *regularization* which allows arbitrarily complex models but penalizes complexity. The third is to combine many overfit models in an ensemble. We'll learn how to use all these techniques in the context of a single dataset.
At the end of this workshop you will know what to do using the validation set when confronted with overfitting, and how to internalize these techniques into your machine learning workflows