BSA: Ball Sparse Attention for Large-scale Geometries

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Brita, Catalin E., Nguyen, Hieu, Chanchu, Lohithsai Yadala, Nagy, Domonkos, Zhdanov, Maksim
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912429932281856
author Brita, Catalin E.
Nguyen, Hieu
Chanchu, Lohithsai Yadala
Nagy, Domonkos
Zhdanov, Maksim
author_facet Brita, Catalin E.
Nguyen, Hieu
Chanchu, Lohithsai Yadala
Nagy, Domonkos
Zhdanov, Maksim
contents Self-attention scales quadratically with input size, limiting its use for large-scale physical systems. Although sparse attention mechanisms provide a viable alternative, they are primarily designed for regular structures such as text or images, making them inapplicable for irregular geometries. In this work, we present Ball Sparse Attention (BSA), which adapts Native Sparse Attention (NSA) (Yuan et al., 2025) to unordered point sets by imposing regularity using the Ball Tree structure from the Erwin Transformer (Zhdanov et al., 2025). We modify NSA's components to work with ball-based neighborhoods, yielding a global receptive field at sub-quadratic cost. On an airflow pressure prediction task, we achieve accuracy comparable to Full Attention while significantly reducing the theoretical computational complexity. Our implementation is available at https://github.com/britacatalin/bsa.
format Preprint
id arxiv_https___arxiv_org_abs_2506_12541
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BSA: Ball Sparse Attention for Large-scale Geometries
Brita, Catalin E.
Nguyen, Hieu
Chanchu, Lohithsai Yadala
Nagy, Domonkos
Zhdanov, Maksim
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Self-attention scales quadratically with input size, limiting its use for large-scale physical systems. Although sparse attention mechanisms provide a viable alternative, they are primarily designed for regular structures such as text or images, making them inapplicable for irregular geometries. In this work, we present Ball Sparse Attention (BSA), which adapts Native Sparse Attention (NSA) (Yuan et al., 2025) to unordered point sets by imposing regularity using the Ball Tree structure from the Erwin Transformer (Zhdanov et al., 2025). We modify NSA's components to work with ball-based neighborhoods, yielding a global receptive field at sub-quadratic cost. On an airflow pressure prediction task, we achieve accuracy comparable to Full Attention while significantly reducing the theoretical computational complexity. Our implementation is available at https://github.com/britacatalin/bsa.
title BSA: Ball Sparse Attention for Large-scale Geometries
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.12541