Beyond AutoML: Pipelines, Trainers, and Transforms

We’ve now covered regression and classification using ML.NET’s AutoML experiments API. In the previous two chapters, we focused more on the task that we wanted to accomplish than the actual process of the machine learning logic, and with good reason – there’s a lot to learn when learning about machine learning.

In this chapter, we’ll lift the veil on what AutoML does by recreating our regression experiment from Chapter 8 without using AutoML at all. This will show us the important role of pipelines, trainers, and transforms in the model training process.

Once we have a firm understanding of ML.NET without AutoML, we’ll see how advanced AutoML experiments can be configured by using AutoML pipelines. We’ll see how this approach gives you more fine-grained control of the training process while still taking advantage of AutoML’s hyperparameter tuning abilities.

Along the...