welcome

To get the most out of these pages, you should be comfortable reading and writing Python or a similar language. We will use PyTorch for the neural material and lean on some statistics, since we are talking about data-driven applications, but nothing past what you'd meet in an introductory course, like probability or taking the average of a series of numbers. If you have had contact with machine learning, that will help, but I have tried to introduce the necessary ideas as we go rather than assume they are known.

I wrote Practical Recommender Systems some years ago, and since then the ground has shifted under this field. Embeddings and transformers are core components rather than fancy additions. Large language models arrived and changed what a recommender can even be asked to do. And the oldest question of all — is this actually worth putting in production? — has somehow gotten harder, rather than easier, to answer. This book is my attempt to make sense of this new landscape without losing the fundamentals that it is still based on.