This chapter covers:
- Understanding the tension between generalization and optimization, the fundamental issue in machine learning
- Evaluation methods for machine learning models
- Best practices to improve model fitting
- Best practices to achieve better generalization
After the three practical examples in chapter 4, you should be starting to feel familiar with how to approach classification and regression problems using neural networks, and you’ve witnessed the central problem of machine learning: overfitting. This chapter will formalize some of your new intuition about machine learning into a solid conceptual framework, highlighting the importance of accurate model evaluation and the balance between training and generalization.
In the three examples presented in chapter 3 — predicting movie reviews, topic classification, and house-price regression — we split the data into a training set, a validation set, and a test set. The reason not to evaluate the models on the same data they were trained on quickly became evident: after just a few epochs, performance on never-before-seen data started diverging from performance on the training data — which always improves as training progresses. The models started to overfit. Overfitting happens in every machine-learning problem.