# Link and code: Fast indexing with graphs and compact regression codes

arXiv: https://arxiv.org/abs/1804.09996 · HTML: https://mlsanity.com/paper/1804.09996/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Apr 2018.

## Built on by (most cited first, 9 of 9)

- [The Faiss library](https://mlsanity.com/paper/2401.08281/) (2024)
- [LeanVec: Searching vectors faster by making them fit](https://mlsanity.com/paper/2312.16335/) (2023)
- [FINGER: Fast Inference for Graph-based Approximate Nearest Neighbor Search](https://mlsanity.com/paper/2206.11408/) (2022)
- [LEANN: A Low-Storage Vector Index](https://mlsanity.com/paper/2506.08276/) (2025)
- [Large-Scale Visual Search with Binary Distributed Graph at Alibaba](https://mlsanity.com/paper/2102.04656/) (2021)
- [Product Re-identification System in Fully Automated Defect Detection](https://mlsanity.com/paper/2112.10324/) (2021)
- [ARM 4-BIT PQ: SIMD-based Acceleration for Approximate Nearest Neighbor Search on ARM](https://mlsanity.com/paper/2203.02505/) (2022)
- [Lossless Compression of Vector IDs for Approximate Nearest Neighbor Search](https://mlsanity.com/paper/2501.10479/) (2025)
- [LotusFilter: Fast Diverse Nearest Neighbor Search via a Learned Cutoff Table](https://mlsanity.com/paper/2506.04790/) (2025)

## Abstract (arXiv)

Similarity search approaches based on graph walks have recently attained outstanding speed-accuracy trade-offs, taking aside the memory requirements. In this paper, we revisit these approaches by considering, additionally, the memory constraint required to index billions of images on a single server. This leads us to propose a method based both on graph traversal and compact representations. We encode the indexed vectors using quantization and exploit the graph structure to refine the similarity estimation. In essence, our method takes the best of these two worlds: the search strategy is based on nested graphs, thereby providing high precision with a relatively small set of comparisons. At the same time it offers a significant memory compression. As a result, our approach outperforms the state of the art on operating points considering 64-128 bytes per vector, as demonstrated by our results on two billion-scale public benchmarks.