Chapter 12 Model tuning and the dangers of overfitting

Learning objectives:

• Recognize examples of tuning parameters.
  • Recognize hyperparameters for machine learning models.
  • Recognize tuning parameters for preprocessing techniques.
  • Recognize structural parameters for classical statistical models.
  • Recognize examples of parameters that should not be tuned.
• Explain how different metrics can lead to different decisions about the choice of tuning parameter values.
• Explain how poor parameter estimates can lead to overfitting of training data.
• Recognize strategies for optimizing tuning parameters.
  • Compare and contrast grid search and iterative search.
• Use tune::tune() and the {dials} package to optimize tuning parameters.