Chapter 3: Feature Store Fundamentals, Terminology, and Usage
In the last chapter, we discussed the need to bring features into production and different ways of doing so, along with a look at common issues with these approaches and how feature stores can solve them. We have built up a lot of expectations about feature stores, and it's time to understand how they work. As mentioned in the last chapter, a feature store is different from a traditional database – it is a data storage service for managing machine learning features, a hybrid system that can be used for storage and retrieval of historical features for model training. It can also serve the latest features at low latency for real-time prediction, and at sub-second latency for batch prediction.
In this chapter, we will discuss what a feature store is, how it works, and the range of terminology used in the feature store world. For this chapter, we will use one of the most widely used open source feature stores...