A Comprehensive Survey on Vector Database

Category: Research Papers & Surveys
Tags: vector-databases, survey, ANNS, architecture
Source: arXiv:2310.11703

Overview

This academic survey provides an in-depth review of vector databases, focusing on their architecture, storage, retrieval techniques, and the challenges they face. Vector databases are specialized systems for storing and retrieving high-dimensional data, a task not well-handled by traditional database management systems.

Features

• Comprehensive Overview: Covers the current landscape of vector database architectures.
• Algorithmic Approaches for ANNS: Categorizes and reviews major approaches to the Approximate Nearest Neighbor Search (ANNS) problem, including:
  • Hash-based methods
  • Tree-based methods
  • Graph-based methods
  • Quantization-based methods
• Challenges: Discusses existing challenges in vector database design and implementation.
• Integration with Large Language Models: Explores how vector databases can be combined with large language models, opening up new application possibilities.
• Foundational Knowledge: Offers valuable insights for understanding and building vector database systems.

Pricing

n/a (This is a research paper freely available on arXiv)

Access

• Read the paper on arXiv