A Comparison between Recursive Neural Networks and Graph Neural Networks

Vincenzo Di Massa; Gabriele Monfardini; Lorenzo Sarti; Franco Scarselli; Marco Maggini; Marco Gori

doi:10.1109/ijcnn.2006.246763

Abstract

Recursive neural networks (RNNs) and graph neural networks (GNNs) are two connectionist models that can directly process graphs. RNNs and GNNs exploit a similar processing framework, but they can be applied to different input domains. RNNs require the input graphs to be directed and acyclic, whereas GNNs can process any kind of graphs. The aim of this paper consists in understanding whether such a difference affects the behaviour of the models on a real application. An experimental comparison on an image classification problem is presented, showing that GNNs outperforms RNNs. Moreover the main differences between the models are also discussed w.r.t. their input domains, their approximation capabilities and their learning algorithms.

Keywords

Recurrent neural networkConnectionismComputer scienceExploitArtificial intelligenceArtificial neural networkGraphDirected acyclic graphProcess (computing)Directed graphMachine learningTheoretical computer scienceAlgorithm

Affiliated Institutions

University of Siena IT

Related Publications

Neural Network for Graphs: A Contextual Constructive Approach

Alessio Micheli

This paper presents a new approach for learning in structured domains (SDs) using a constructive neural network for graphs (NN4G). The new model allows the extension of the inpu...

2009 IEEE Transactions on Neural Networks 607 citations

The Graph Neural Network Model

Franco Scarselli , M. Gori , Ah Chung Tsoi +2 more

Many underlying relationships among data in several areas of science and engineering, e.g., computer vision, molecular chemistry, molecular biology, pattern recognition, and dat...

2008 IEEE Transactions on Neural Networks 8377 citations

Learning task-dependent distributed representations by backpropagation through structure

Christoph Goller , Andreas Küchler

While neural networks are very successfully applied to the processing of fixed-length vectors and variable-length sequences, the current state of the art does not allow the effi...

2002 Proceedings of International Conferen... 591 citations

Spectral Networks and Locally Connected Networks on Graphs

Joan Bruna , Wojciech Zaremba , Arthur Szlam +1 more

Convolutional Neural Networks are extremely efficient architectures in image and audio recognition tasks, thanks to their ability to exploit the local translational invariance o...

2013 arXiv (Cornell University) 2710 citations

A self-organizing map for adaptive processing of structured data

Markus Hagenbuchner , Alessandro Sperduti , Ah Chung Tsoi

Recent developments in the area of neural networks produced models capable of dealing with structured data. Here, we propose the first fully unsupervised model, namely an extens...

2003 IEEE Transactions on Neural Networks 180 citations

Publication Info

Year: 2006
Type: article
Pages: 778-785
Citations: 39
Access: Closed

External Links

View on DOI.org

Social Impact

Altmetric

A Comparison between Recursive Neural Networks and Graph Neural Networks

PlumX Metrics

Social media, news, blog, policy document mentions

Citation Metrics

OpenAlex

Cite This

APA Style

                            
                                    Vincenzo Di Massa, 
                                
                                    Gabriele Monfardini, 
                                
                                    Lorenzo Sarti
                                
                                et al.
                            
                            (2006). 
                            A Comparison between Recursive Neural Networks and Graph Neural Networks. 
                            The 2006 IEEE International Joint Conference on Neural Network Proceedings
                            
                            , 778-785.
                            https://doi.org/10.1109/ijcnn.2006.246763

Identifiers

DOI: 10.1109/ijcnn.2006.246763