If we analyze each of the examples presented so far, we see that each follows a similar pattern. First is the definition of the problem or desired functionality. Once we have established what we want to do, we then identify the available data and/or what data is required. With the data in hand, our next step is to create our ML model and prepare the data for training.

After training, something we hadn't discussed here, is validating our ML model, that is, testing that it satisfactorily achieves what we require of it. An example is being able to make an accurate prediction. Once we have trained a model, we can make use of it by feeding in real data, that is, data outside our training set. In the following diagram, we see these steps summarized for training and inference: 

We will spend most of our time using trained models in this book, but understanding how we arrive at these models will prove helpful as you start creating your own intelligent apps. This will also help you identify opportunities to apply ML on existing data or inspire you to seek out new data sources. It's also worth noting that the preprocessing step on training data is equivalent to preprocessing on input data when performing inference—something we will spend a lot of time discussing and coding for throughout this book.