Filters








21,210 Hits in 3.7 sec

A Fair Comparison of Graph Neural Networks for Graph Classification [article]

Federico Errica, Marco Podda, Davide Bacciu, Alessio Micheli
2022 arXiv   pre-print
We believe that this work can contribute to the development of the graph learning field, by providing a much needed grounding for rigorous evaluations of graph classification models.  ...  As such, several Graph Neural Network models have been developed to effectively tackle graph classification. However, experimental procedures often lack rigorousness and are hardly reproducible.  ...  RELATED WORK Graph Neural Networks At the core of GNNs is the idea to compute a state for each node in a graph, which is iteratively updated according to the state of neighboring nodes.  ... 
arXiv:1912.09893v3 fatcat:h4ngcck57nc3dj6yor646vgz7q

Say No to the Discrimination: Learning Fair Graph Neural Networks with Limited Sensitive Attribute Information [article]

Enyan Dai, Suhang Wang
2021 arXiv   pre-print
Graph neural networks (GNNs) have shown great power in modeling graph structured data.  ...  Though extensive studies of fair classification have been conducted on i.i.d data, methods to address the problem of discrimination on non-i.i.d data are rather limited.  ...  Advanced Institute of Technology under grant #225003.  ... 
arXiv:2009.01454v5 fatcat:wb2bqq4khfcxfim4k7h2fxkpou

Benchmarking Graph Neural Networks on Link Prediction [article]

Xing Wang, Alexander Vinel
2021 arXiv   pre-print
In this paper, we benchmark several existing graph neural network (GNN) models on different datasets for link predictions.  ...  analysis are performed, and results from several different papers are replicated, also a more fair and systematic comparison are provided.  ...  Cora, CiteSeer, and PubMed are commonly used benchmark datasets for tasks related to graph neural networks such as link prediction, node classification, and graph classification.  ... 
arXiv:2102.12557v1 fatcat:jbk4pekrk5cdrmwrz3tqlkmlgu

FairEdit: Preserving Fairness in Graph Neural Networks through Greedy Graph Editing [article]

Donald Loveland, Jiayi Pan, Aaresh Farrokh Bhathena, Yiyang Lu
2022 arXiv   pre-print
Graph Neural Networks (GNNs) have proven to excel in predictive modeling tasks where the underlying data is a graph.  ...  FairEdit performs efficient edge editing by leveraging gradient information of a fairness loss to find edges that improve fairness.  ...  While fairness has been studied for traditional machine learning algorithms and even deep neural networks, graph neural networks (GNNs), neural networks that operate directly on graphs, have only received  ... 
arXiv:2201.03681v3 fatcat:wmqrd3e7qbeoxglzjmncx4wydm

Mutual Information Maximization in Graph Neural Networks [article]

Xinhan Di, Pengqian Yu, Rui Bu, Mingchao Sun
2020 arXiv   pre-print
A variety of graph neural networks (GNNs) frameworks for representation learning on graphs have been recently developed.  ...  Based on a series of experiments conducted on several benchmark datasets, we show that the proposed approach improves the state-of-the-art performance for four types of graph tasks, including supervised  ...  For a fair comparison, we keep all parameters the same as those in the baseline models GCN [2] , GIN [17] , KNN-LDS [22] , graph Markov neural network (GMNN) [19] , Graphite [20] , Graph-Unet [21  ... 
arXiv:1905.08509v4 fatcat:pj3zvlsykndjflw6louwy3bw44

A Variational Graph Autoencoder for Manipulation Action Recognition and Prediction [article]

Gamze Akyol, Sanem Sariel, Eren Erdal Aksoy
2021 arXiv   pre-print
Our network has a variational autoencoder structure with two branches: one for identifying the input graph type and one for predicting the future graphs.  ...  The input of the proposed network is a set of semantic graphs which store the spatial relations between subjects and objects in the scene.  ...  In the MANIAC dataset, 4 consecutive scene graphs are merged into one to be fed to the network as an input, whereas graphs are treated individually in MSRC-9 to have a fair comparison with other state-of-the-art  ... 
arXiv:2110.13280v1 fatcat:6ywfdrextrc3bd32jjvsdhyqbm

Knowledge-Embedded Representation Learning for Fine-Grained Image Recognition

Tianshui Chen, Liang Lin, Riquan Chen, Yang Wu, Xiaonan Luo
2018 Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence  
Specifically, we organize the rich visual concepts in the form of knowledge graph and employ a Gated Graph Neural Network to propagate node message through the graph for generating the knowledge representation  ...  In this work, we investigate how to unify rich professional knowledge with deep neural network architectures and propose a Knowledge-Embedded Representation Learning (KERL) framework for handling the problem  ...  Specifically, we organize the rich visual concepts in the form of knowledge graph and employ a Gated Graph Neural Network to propagate node message through the graph for generating the knowledge representation  ... 
doi:10.24963/ijcai.2018/87 dblp:conf/ijcai/ChenLCWL18 fatcat:iek67dt6mzftld5cll247t6mfa

NF-GNN: Network Flow Graph Neural Networks for Malware Detection and Classification [article]

Julian Busch, Anton Kocheturov, Volker Tresp, Thomas Seidl
2021 arXiv   pre-print
Our approach first extracts flow graphs and subsequently classifies them using a novel graph neural network model.  ...  Experiments on four different prediction tasks consistently demonstrate the advantages of our approach and show that our graph neural network model can boost detection performance by a significant margin  ...  A HYPER-PARAMETER OPTIMIZATION For the sake of a fair comparison, hyper-parameters of all models are optimized by a grid search based on performance on a separate validation set.  ... 
arXiv:2103.03939v2 fatcat:wjgxdcqvybbtbiqraariaixa6u

Hypergraph Convolution and Hypergraph Attention [article]

Song Bai, Feihu Zhang, Philip H.S. Torr
2020 arXiv   pre-print
With the two operators, a graph neural network is readily extended to a more flexible model and applied to diverse applications where non-pairwise relationships are observed.  ...  Recently, graph neural networks have attracted great attention and achieved prominent performance in various research fields.  ...  The comparison is primarily done with Graph Convolution Network (GCN) [22] and Graph Attention Network (GAT) [31] , which are two latest representatives of graph neural networks that have close relationships  ... 
arXiv:1901.08150v2 fatcat:gqanvg6tqrhchlx2dlub5iosgu

Dynamic Graph: Learning Instance-aware Connectivity for Neural Networks [article]

Kun Yuan, Quanquan Li, Dapeng Chen, Aojun Zhou, Junjie Yan
2020 arXiv   pre-print
We generate edge weights by a learnable module router and select the edges whose weights are larger than a threshold, to adjust the connectivity of the neural network structure.  ...  One practice of employing deep neural networks is to apply the same architecture to all the input instances.  ...  ., 2018; , for fair comparisons.  ... 
arXiv:2010.01097v1 fatcat:gndgp47sgbbb7fhun6pkkyijxe

Describe Molecules by a Heterogeneous Graph Neural Network with Transformer-like Attention for Supervised Property Predictions

Daiguo Deng, Zengrong Lei, Xiaobin Hong, Ruochi Zhang, Fengfeng Zhou
2022 ACS Omega  
This study represented a molecule using a heterogeneous graph neural network (MolHGT), in which there were different types of nodes and different types of edges.  ...  The rapid development of natural language processing and graph neural network (GNN) further pushed the state-of-the-art prediction performance of molecular property to a new level.  ...  We created a new molecular heterogeneity setting for the graph neural network, including 11 types of nodes and 4 types of chemical bonds.  ... 
doi:10.1021/acsomega.1c06389 pmid:35128279 pmcid:PMC8811943 fatcat:bslay4atnjg6tnursq4mzgowju

Quaternion Graph Neural Networks [article]

Dai Quoc Nguyen and Tu Dinh Nguyen and Dinh Phung
2021 arXiv   pre-print
Our QGNN obtains state-of-the-art results on a range of benchmark datasets for graph classification and node classification.  ...  Recently, graph neural networks (GNNs) have become an important and active research direction in deep learning.  ...  We report the new results of TextING for the model which obtains the highest accuracy on the validation set for a fair comparison.  ... 
arXiv:2008.05089v6 fatcat:sffmebl6qvbm5imef7p2vbvkyy

Fairness-Aware Node Representation Learning [article]

Öykü Deniz Köse, Yanning Shen
2021 arXiv   pre-print
classification accuracy to state-of-the-art contrastive methods for node classification.  ...  Particularly, recent developments in contrastive learning have led to promising results in unsupervised node representation learning for a number of tasks.  ...  Node embeddings are obtained using a two-layer graph convolutional network (GCN) [4] for all GNN-based baselines, which is kept the same as the one used in [17, 30] to ensure fair comparison.  ... 
arXiv:2106.05391v1 fatcat:gbzvlecaeresdjn52ec3osvdwe

Population Graph-Based Multi-Model Ensemble Method for Diagnosing Autism Spectrum Disorder

Zarina Rakhimberdina, Xin Liu, and Tsuyoshi Murata
2020 Sensors  
Subsequently, we utilize a neural network architecture to combine multiple graph-based models.  ...  Recently, graph neural networks have found increasing application in domains where the population's structure is modeled as a graph.  ...  Recently, Graph Convolutional Neural Network, or GCN, was introduced as an efficient method for node classification on graphs [23, 24] .  ... 
doi:10.3390/s20216001 pmid:33105909 pmcid:PMC7660214 fatcat:cybol32tl5gb7jl7wvwd5adsgy

Efficient Deep Feature Learning and Extraction via StochasticNets

Mohammad Javad Shafiee, Parthipan Siva, Paul Fieguth, Alexander Wong
2016 2016 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
Deep neural networks are a powerful tool for feature learning and extraction.  ...  classification accuracy than conventional deep neural networks.  ...  While one can form practically any type of deep neural network as a realization of random graphs, an important design consideration for forming deep neural networks as random graph realizations is that  ... 
doi:10.1109/cvprw.2016.141 dblp:conf/cvpr/ShafieeSFW16a fatcat:g5h3jil4vnd3lbemg3nzux4gz4
« Previous Showing results 1 — 15 out of 21,210 results