Billion-scale Similarity Search Using a Hybrid Indexing Approach with Advanced Filtering

TL;DR

This paper introduces a hybrid indexing method that enhances billion-scale similarity search by integrating multi-dimensional filters into the IVF-Flat structure, optimized for CPU inference and large datasets.

Contribution

It extends classical IVF-Flat indexes with multi-dimensional filtering, enabling efficient high-dimensional similarity search on large datasets using CPU-based systems.

Findings

Effective filtering on billion-scale datasets

Fast retrieval in high-dimensional spaces

Cost-efficient CPU-based large-scale search

Abstract

This paper presents a novel approach for similarity search with complex filtering capabilities on billion-scale datasets, optimized for CPU inference. Our method extends the classical IVF-Flat index structure to integrate multi-dimensional filters. The proposed algorithm combines dense embeddings with discrete filtering attributes, enabling fast retrieval in high-dimensional spaces. Designed specifically for CPU-based systems, our disk-based approach offers a cost-effective solution for large-scale similarity search. We demonstrate the effectiveness of our method through a case study, showcasing its potential for various practical uses.