Graph Neural Network Generalization with Gaussian Mixture Model Based Augmentation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Abbahaddou, Yassine, Malliaros, Fragkiskos D., Lutzeyer, Johannes F., Aboussalah, Amine Mohamed, Vazirgiannis, Michalis
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909639923204096
author Abbahaddou, Yassine
Malliaros, Fragkiskos D.
Lutzeyer, Johannes F.
Aboussalah, Amine Mohamed
Vazirgiannis, Michalis
author_facet Abbahaddou, Yassine
Malliaros, Fragkiskos D.
Lutzeyer, Johannes F.
Aboussalah, Amine Mohamed
Vazirgiannis, Michalis
contents Graph Neural Networks (GNNs) have shown great promise in tasks like node and graph classification, but they often struggle to generalize, particularly to unseen or out-of-distribution (OOD) data. These challenges are exacerbated when training data is limited in size or diversity. To address these issues, we introduce a theoretical framework using Rademacher complexity to compute a regret bound on the generalization error and then characterize the effect of data augmentation. This framework informs the design of GRATIN, an efficient graph data augmentation algorithm leveraging the capability of Gaussian Mixture Models (GMMs) to approximate any distribution. Our approach not only outperforms existing augmentation techniques in terms of generalization but also offers improved time complexity, making it highly suitable for real-world applications.
format Preprint
id arxiv_https___arxiv_org_abs_2411_08638
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Graph Neural Network Generalization with Gaussian Mixture Model Based Augmentation
Abbahaddou, Yassine
Malliaros, Fragkiskos D.
Lutzeyer, Johannes F.
Aboussalah, Amine Mohamed
Vazirgiannis, Michalis
Machine Learning
Social and Information Networks
Applications
Graph Neural Networks (GNNs) have shown great promise in tasks like node and graph classification, but they often struggle to generalize, particularly to unseen or out-of-distribution (OOD) data. These challenges are exacerbated when training data is limited in size or diversity. To address these issues, we introduce a theoretical framework using Rademacher complexity to compute a regret bound on the generalization error and then characterize the effect of data augmentation. This framework informs the design of GRATIN, an efficient graph data augmentation algorithm leveraging the capability of Gaussian Mixture Models (GMMs) to approximate any distribution. Our approach not only outperforms existing augmentation techniques in terms of generalization but also offers improved time complexity, making it highly suitable for real-world applications.
title Graph Neural Network Generalization with Gaussian Mixture Model Based Augmentation
topic Machine Learning
Social and Information Networks
Applications
url https://arxiv.org/abs/2411.08638