The neural network approach to creating a recommendation system

In recent years, neural networks have been the Swiss army knife when it comes to solving many problems in the field of machine learning (ML). This is evident in areas of ML breakthroughs such as image classification/segmentation and natural language processing. With the availability of data, neural networks have been successfully used to build large-scale recommendation systems such as the ones used at Netflix (https://research.netflix.com/research-area/machine-learning) and YouTube (https://research.google/pubs/pub45530/).

Although there are different approaches to building a recommendation system with neural networks, they all rely on one major fact: they need an efficient way to learn similarities between items or users. In this chapter, we'll leverage a concept called embeddings to efficiently learn these similarities so that we can easily power our recommendation system.

But first, what are embeddings...