Attending to Topological Spaces: The Cellular Transformer

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ballester, Rubén, Hernández-García, Pablo, Papillon, Mathilde, Battiloro, Claudio, Miolane, Nina, Birdal, Tolga, Casacuberta, Carles, Escalera, Sergio, Hajij, Mustafa
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910459795341312
author Ballester, Rubén
Hernández-García, Pablo
Papillon, Mathilde
Battiloro, Claudio
Miolane, Nina
Birdal, Tolga
Casacuberta, Carles
Escalera, Sergio
Hajij, Mustafa
author_facet Ballester, Rubén
Hernández-García, Pablo
Papillon, Mathilde
Battiloro, Claudio
Miolane, Nina
Birdal, Tolga
Casacuberta, Carles
Escalera, Sergio
Hajij, Mustafa
contents Topological Deep Learning seeks to enhance the predictive performance of neural network models by harnessing topological structures in input data. Topological neural networks operate on spaces such as cell complexes and hypergraphs, that can be seen as generalizations of graphs. In this work, we introduce the Cellular Transformer (CT), a novel architecture that generalizes graph-based transformers to cell complexes. First, we propose a new formulation of the usual self- and cross-attention mechanisms, tailored to leverage incidence relations in cell complexes, e.g., edge-face and node-edge relations. Additionally, we propose a set of topological positional encodings specifically designed for cell complexes. By transforming three graph datasets into cell complex datasets, our experiments reveal that CT not only achieves state-of-the-art performance, but it does so without the need for more complex enhancements such as virtual nodes, in-domain structural encodings, or graph rewiring.
format Preprint
id arxiv_https___arxiv_org_abs_2405_14094
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Attending to Topological Spaces: The Cellular Transformer
Ballester, Rubén
Hernández-García, Pablo
Papillon, Mathilde
Battiloro, Claudio
Miolane, Nina
Birdal, Tolga
Casacuberta, Carles
Escalera, Sergio
Hajij, Mustafa
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Algebraic Topology
Topological Deep Learning seeks to enhance the predictive performance of neural network models by harnessing topological structures in input data. Topological neural networks operate on spaces such as cell complexes and hypergraphs, that can be seen as generalizations of graphs. In this work, we introduce the Cellular Transformer (CT), a novel architecture that generalizes graph-based transformers to cell complexes. First, we propose a new formulation of the usual self- and cross-attention mechanisms, tailored to leverage incidence relations in cell complexes, e.g., edge-face and node-edge relations. Additionally, we propose a set of topological positional encodings specifically designed for cell complexes. By transforming three graph datasets into cell complex datasets, our experiments reveal that CT not only achieves state-of-the-art performance, but it does so without the need for more complex enhancements such as virtual nodes, in-domain structural encodings, or graph rewiring.
title Attending to Topological Spaces: The Cellular Transformer
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Algebraic Topology
url https://arxiv.org/abs/2405.14094