Attic

The Attic collects specialized and background topics that are valuable but sit outside the main narrative arc. It leads with Gaussian processes, the Bayesian-nonparametric background that the next chapter builds on — a principled way to reason about functions and uncertainty. Hyperparameter optimization then turns that reasoning into practice, treating the search for good training configurations as an optimization problem in its own right, from random search to Bayesian and multi-fidelity methods.

Finally, recommender systems apply representation learning to a domain with its own conventions — implicit feedback, matrix factorization, and ranking — showing how the book’s tools transfer well beyond the vision-and-language mainstream. These chapters are self-contained references rather than prerequisites for the rest of the book.