Vector Database: History and Basic Concept

An efficient solution to storing embedding vectors

MAKSYM / UNSPLASH.COM / DATABASE

In the last 3 years, the term "Vector Database" has gained attention, particularly in the realm of AI advancement. Despite its recent prominence, vectors have long been integrated into machine learning and AI, often referred to as "Embedding." The fundamental concept behind a vector database is to enhance the management of vectors, particularly for querying similarity. This capability significantly enhances AI's ability to efficiently and accurately retrieve knowledge and context.

Introduction

Vector DB has indeed traversed a significant trajectory, rooted in the utilization of vector space models in distributional semantics, dating back to the 1990s or even earlier. Over time, numerous algorithms have been developed to effectively convert any digital entity into a vector representation. Moreover, an array of algorithms has also been crafted to streamline the processes of indexing and querying within the resultant vector space. In the forthcoming discourse, attention will be directed towards how the vector becomes a vector database based on that trajectory and how it works in indexing and querying.