Saved in:
Bibliographic Details
Main Authors: Nousias, Akis, Nousias, Stavros
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.01839
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918234108723200
author Nousias, Akis
Nousias, Stavros
author_facet Nousias, Akis
Nousias, Stavros
contents Currently, prominent Transformer architectures applied on graphs and meshes for shape analysis tasks employ traditional attention layers that heavily utilize spectral features requiring costly eigenvalue decomposition-based methods. To encode the mesh structure, these methods derive positional embeddings, that heavily rely on eigenvalue decomposition based operations, e.g. on the Laplacian matrix, or on heat-kernel signatures, which are then concatenated to the input features. This paper proposes a novel approach inspired by the explicit construction of the Hodge Laplacian operator in Discrete Exterior Calculus as a product of discrete Hodge operators and exterior derivatives, i.e. $(L := \star_0^{-1} d_0^T \star_1 d_0)$. We adjust the Transformer architecture in a novel deep learning layer that utilizes the multi-head attention mechanism to approximate Hodge matrices $\star_0$, $\star_1$ and $\star_2$ and learn families of discrete operators $L$ that act on mesh vertices, edges and faces. Our approach results in a computationally-efficient architecture that achieves comparable performance in mesh segmentation and classification tasks, through a direct learning framework, while eliminating the need for costly eigenvalue decomposition operations or complex preprocessing operations.
format Preprint
id arxiv_https___arxiv_org_abs_2509_01839
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle HodgeFormer: Transformers for Learnable Operators on Triangular Meshes through Data-Driven Hodge Matrices
Nousias, Akis
Nousias, Stavros
Graphics
Artificial Intelligence
Computer Vision and Pattern Recognition
Currently, prominent Transformer architectures applied on graphs and meshes for shape analysis tasks employ traditional attention layers that heavily utilize spectral features requiring costly eigenvalue decomposition-based methods. To encode the mesh structure, these methods derive positional embeddings, that heavily rely on eigenvalue decomposition based operations, e.g. on the Laplacian matrix, or on heat-kernel signatures, which are then concatenated to the input features. This paper proposes a novel approach inspired by the explicit construction of the Hodge Laplacian operator in Discrete Exterior Calculus as a product of discrete Hodge operators and exterior derivatives, i.e. $(L := \star_0^{-1} d_0^T \star_1 d_0)$. We adjust the Transformer architecture in a novel deep learning layer that utilizes the multi-head attention mechanism to approximate Hodge matrices $\star_0$, $\star_1$ and $\star_2$ and learn families of discrete operators $L$ that act on mesh vertices, edges and faces. Our approach results in a computationally-efficient architecture that achieves comparable performance in mesh segmentation and classification tasks, through a direct learning framework, while eliminating the need for costly eigenvalue decomposition operations or complex preprocessing operations.
title HodgeFormer: Transformers for Learnable Operators on Triangular Meshes through Data-Driven Hodge Matrices
topic Graphics
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.01839