Chapter 3: Automating the Machine Learning Pipeline with AutoKeras
Automating the machine learning pipeline involves automating a series of processes such as data exploration, data preprocessing, feature engineering, algorithm selection, model training, and hyperparameter tuning.
This chapter explains the standard machine learning pipeline and how to automate some of them with AutoKeras. We will also describe the main data preparation best practices to apply before training a model. The post-data preparation steps are performed by AutoKeras and we will see them in depth in later chapters.
As we saw in the first chapter, AutoKeras can automate all pipeline modeling steps by applying hyperparameter optimization and Neural Architecture Search (NAS), but some data preprocessing before these steps must be done by hand or with other tools.
We will explain the data representations expected by our model, as well as the basic preprocessing techniques that AutoKeras applies. By the...
