Saved in:
Bibliographic Details
Main Authors: Rohatgi, Dhruv, Marwah, Tanya, Lipton, Zachary Chase, Lu, Jianfeng, Moitra, Ankur, Risteski, Andrej
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.09867
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917801891987456
author Rohatgi, Dhruv
Marwah, Tanya
Lipton, Zachary Chase
Lu, Jianfeng
Moitra, Ankur
Risteski, Andrej
author_facet Rohatgi, Dhruv
Marwah, Tanya
Lipton, Zachary Chase
Lu, Jianfeng
Moitra, Ankur
Risteski, Andrej
contents Graph neural networks (GNNs) are the dominant approach to solving machine learning problems defined over graphs. Despite much theoretical and empirical work in recent years, our understanding of finer-grained aspects of architectural design for GNNs remains impoverished. In this paper, we consider the benefits of architectures that maintain and update edge embeddings. On the theoretical front, under a suitable computational abstraction for a layer in the model, as well as memory constraints on the embeddings, we show that there are natural tasks on graphical models for which architectures leveraging edge embeddings can be much shallower. Our techniques are inspired by results on time-space tradeoffs in theoretical computer science. Empirically, we show architectures that maintain edge embeddings almost always improve on their node-based counterparts -- frequently significantly so in topologies that have ``hub'' nodes.
format Preprint
id arxiv_https___arxiv_org_abs_2410_09867
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Towards characterizing the value of edge embeddings in Graph Neural Networks
Rohatgi, Dhruv
Marwah, Tanya
Lipton, Zachary Chase
Lu, Jianfeng
Moitra, Ankur
Risteski, Andrej
Machine Learning
Graph neural networks (GNNs) are the dominant approach to solving machine learning problems defined over graphs. Despite much theoretical and empirical work in recent years, our understanding of finer-grained aspects of architectural design for GNNs remains impoverished. In this paper, we consider the benefits of architectures that maintain and update edge embeddings. On the theoretical front, under a suitable computational abstraction for a layer in the model, as well as memory constraints on the embeddings, we show that there are natural tasks on graphical models for which architectures leveraging edge embeddings can be much shallower. Our techniques are inspired by results on time-space tradeoffs in theoretical computer science. Empirically, we show architectures that maintain edge embeddings almost always improve on their node-based counterparts -- frequently significantly so in topologies that have ``hub'' nodes.
title Towards characterizing the value of edge embeddings in Graph Neural Networks
topic Machine Learning
url https://arxiv.org/abs/2410.09867