A vector is an ordered collection of numerical components. In recommendation systems, vectors represent items, users, or their preferences; dense embedding vectors compress information while preserving meaningful similarity. section 3.1.1 section 3.3.1
Figure 3.2 The movie matrix can have different types of representations; here, a TFIDF and an embedding representation.

Compare a sparse TF-IDF representation with a dense embedding of the same movie information. section 3.1.1
Vector closeness can represent similarity. Cosine similarity compares direction; the dot product reflects direction and magnitude. For unit-length vectors, the two measures are equivalent. section 3.1.2 section 3.4.2
Estimating a movie rating
Matrix factorization represents a user and a movie as vectors in a shared latent space. Their dot product estimates the rating; Peter’s vectors produce 0.90 for Pulp Fiction and −0.18 for Toy Story. section 3.3.1


