What is vector search?

Vector search is a technique for finding similar items within vast amounts of unstructured data. As shown in Figure 4.1, both the data and the search query are transformed into mathematical representations known as vectors, often referred to as embeddings in the context of machine learning applications.

Embeddings are fixed-length sequences of real numbers that encapsulate the meaning and relationships present within the data. This enables searches to be performed based on semantic similarity, going beyond simple keyword matching.

Figure 4.1: Embedding of unstructured data

The similarity between two embeddings is calculated using a distance metric, such as Euclidean distance or cosine similarity. Embeddings that are more similar have a smaller distance between them – in other words, they’re closer together in the embedding space.

Now that you know what vector search is, let’s see different types of architectures...