GriSPy: A Python package for Fixed-Radius Nearest Neighbors Search

TL;DR

GriSPy is a Python package that offers a fast, grid-based method for fixed-radius nearest neighbor searches in high-dimensional data, supporting various metrics and large datasets.

Contribution

It introduces a new regular grid search algorithm optimized for quick fixed-radius nearest neighbor queries in Python, with support for multiple metrics and custom functions.

Findings

Provides three query types: bubble, shell, nth-nearest

Supports Euclidean, Haversine, and Vincenty distance metrics

Efficient for large datasets where brute-force is impractical

Abstract

We present a new regular grid search algorithm for quick fixed-radius nearest-neighbor lookup developed in Python. This module indexes a set of k-dimensional points in a regular grid, with optional periodic conditions, providing a fast approach for nearest neighbors queries. In this first installment we provide three types of queries: $bubble$, $shell$ and the $nth-nearest$; as well as three different metrics of interest in astronomy: the $euclidean$ and two distance functions in spherical coordinates of varying precision, $haversine$ and $Vincenty$; and the possibility of providing a custom distance function. This package results particularly useful for large datasets where a brute-force search turns impractical.