General and Practical Tuning Method for Off-the-Shelf Graph-Based Index: SISAP Indexing Challenge Report by Team UTokyo

TL;DR

This paper presents a universal tuning method for off-the-shelf graph-based indexes in ANN searches, optimizing parameters to improve recall and speed, demonstrated through the SISAP 2023 challenge.

Contribution

Introduces a black-box optimization-based tuning approach for graph indexes, applicable across various conditions and outperforming brute-force tuning.

Findings

Achieved second place in SISAP 2023 Indexing Challenge.

Substantially improved performance over brute-force tuning methods.

Validated the method's broad applicability beyond specific datasets.

Abstract

Despite the efficacy of graph-based algorithms for Approximate Nearest Neighbor (ANN) searches, the optimal tuning of such systems remains unclear. This study introduces a method to tune the performance of off-the-shelf graph-based indexes, focusing on the dimension of vectors, database size, and entry points of graph traversal. We utilize a black-box optimization algorithm to perform integrated tuning to meet the required levels of recall and Queries Per Second (QPS). We applied our approach to Task A of the SISAP 2023 Indexing Challenge and got second place in the 10M and 30M tracks. It improves performance substantially compared to brute force methods. This research offers a universally applicable tuning method for graph-based indexes, extending beyond the specific conditions of the competition to broader uses.