Summary

In this chapter, we studied the statistical model, the learning curve, and curve fitting. We also studied two cultures that Leo Breiman introduced, which describe that any analysis needs data. We went through the different types of training, development, and test data, including their sizes. We studied regularization, which explains what overfitting means in machine learning modeling.

This chapter also explained cross validation and model selection, the 0.632 rule in bootstrapping, and also ROC and AUC in depth.

In the next chapter, we will study evaluating kernel learning, which is the most widely used approach in machine learning.