Chapter 8. Underfitting, overfitting, and the universal workflow of machine learning
This chapter covers
- Why it is important to visualize the model-training process and what the important things are to look for
- How to visualize and understand underfitting and overfitting
- The primary way of dealing with overfitting: regularization, and how to visualize its effect
- What the universal workflow of machine learning is, what steps it includes, and why it is an important recipe that guides all supervised machine-learning tasks
In the previous chapter, you learned how to use tfjs-vis to visualize data before you start designing and training machine-learning models for it. This chapter will start where that one left off and describe how tfjs-vis can be used to visualize the structure and metrics of models during their training. The most important goal in doing so is to spot the all-important phenomena of underfitting and overfitting. Once we can spot them, we’ll delve into how to remedy them and how to verify that our remedying approaches are working using visualization.