Multitask Learning on Graph Neural Networks: Learning Multiple Graph Centrality Measures with a Unified Network

Avelar, Pedro H. C.; Lemos, Henrique; Prates, Marcelo O. R.; Lamb, Luís C.

doi:10.1007/978-3-030-30493-5_63

Cited by 10 publications

(9 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several datasets were used in this work, including a synthetic dataset we built named RGG, inspired by [13]. This dataset was created using graphs generated by five random graph generators available in the NetworkX [14] Python library: Erdos-Renyi (GNP), Random power law tree, Connected Watts-Strogatz smallworld, Holme-Kim and Barabási-Albert.…”

Section: Experiments and Resultsmentioning

confidence: 99%

A bag of nodes primer on weightless graph classification

Barbosa¹,

Carvalho²,

Lima³

et al. 2021

ESANN 2021 Proceedings

View full text Add to dashboard Cite

This paper proposes a weightless architecture for graph classification scenarios. This architecture is a three-headed arrangement composed of graph hand-picked features, a quantization method and a final classifier. Although multiple new strategies for graph classification have been proposed in recent years, it is still necessary to settle comparable studies with respect to weightless neural networks. The proposed architecture is evaluated along with other baseline classifiers and independent strategies, showing that weightless architectures are able to compete with other well-established methods such as graph kernels.

show abstract

Section: Experiments and Resultsmentioning

confidence: 99%

A bag of nodes primer on weightless graph classification

Barbosa¹,

Carvalho²,

Lima³

et al. 2021

ESANN 2021 Proceedings

View full text Add to dashboard Cite

show abstract

“…Holtz et al (2019) and Xie et al (2020) propose multi-task models for concurrently performing node and graph classification. Finally, Avelar et al (2019) introduce a multi-head GNN for learning multiple graph centrality measures, and Li and Ji (2019) propose a MTL method for the extraction of multiple biomedical relations. Other related work includes (Haonan et al, 2019) which introduces a model that can be trained for several tasks singularly, hence, unlike the previously mentioned approaches and our proposed method, it can not perform multiple tasks concurrently.…”

Section: Related Workmentioning

confidence: 99%

Graph Representation Learning for Multi-Task Settings: a Meta-Learning Approach

Buffelli¹,

Vandin²

2022

Preprint

View full text Add to dashboard Cite

Graph Neural Networks (GNNs) have become the state-of-the-art method for many applications on graph structured data. GNNs are a framework for graph representation learning, where a model learns to generate low dimensional node embeddings that encapsulate structural and feature-related information. GNNs are usually trained in an end-to-end fashion, leading to highly specialized node embeddings. While this approach achieves great results in the single-task setting, generating node embeddings that can be used to perform multiple tasks (with performance comparable to single-task models) is still an open problem. We propose a novel training strategy for graph representation learning, based on metalearning, which allows the training of a GNN model capable of producing multi-task node embeddings. Our method avoids the difficulties arising when learning to perform multiple tasks concurrently by, instead, learning to quickly (i.e. with a few steps of gradient descent) adapt to multiple tasks singularly. We show that the embeddings produced by a model trained with our method can be used to perform multiple tasks with comparable or, surprisingly, even higher performance than both single-task and multi-task end-to-end models.

show abstract

“…Firstly, the analysis of schedule networks can identify the tasks' centrality in the schedule network, which can help managers adjust the resources' allocation. Secondly, the parameters of schedule networks are significant in the Graph Neural Network (GNN), which will benefit further work in machine learning or deep learning [58]. The results of the schedule network analysis will be stored in the Pset_ScheduleNetworkAnalysis.…”

Section: Schedule Network Analysismentioning

confidence: 99%

IFC-Based 4D Construction Management Information Model of Prefabricated Buildings and Its Application in Graph Database

et al. 2021

View full text Add to dashboard Cite

Effective data interoperability and schedule analysis play a significant role in improving the management of prefabricated buildings. However, there is a lack of efficient strategies and comprehensive approaches for data interoperability and data-based automated schedule analysis. This paper intends to promote prefabricated buildings’ management by solving these two problems via developing an IFC-based framework consisting of three parts. Firstly, this framework proposed a mechanism to establish an IFC-based 4D construction management information model of prefabricated buildings. Furthermore, a non-relational database—graph database—is introduced to twin this model into a task-centered network to realize the interoperation of construction information among different participants. Finally, graph database-based strategies to update data, automatically analyze construction schedules and visualize the 4D construction management information model are described. The proposed framework is validated in a prefabricated engineering case. In this case, an IFC-based and graph database-based 4D construction management information model is established through IFC standard’s extension. The graph database-based analysis of the model automatically recognizes the engineering case’s critical path information, delay analysis information, and schedule network analysis information. It is illustrated that this framework can successfully establish a unified IFC-based information model of prefabricated buildings’ construction management to prompt effective data interoperability. In addition, the application of this IFC-based information model in graph database can automatically analyze the construction schedules to prevent possible delays in advance. In short, the significance of this paper is to innovatively propose an IFC-based and graph data-based information model to solve the difficulties of ineffective data interoperation and unautomated schedule analysis in prefabricated buildings’ construction management. This study can be the digital foundation of further IFC-based digital twin.

show abstract

Multitask Learning on Graph Neural Networks: Learning Multiple Graph Centrality Measures with a Unified Network

Cited by 10 publications

References 29 publications

A bag of nodes primer on weightless graph classification

A bag of nodes primer on weightless graph classification

Graph Representation Learning for Multi-Task Settings: a Meta-Learning Approach

IFC-Based 4D Construction Management Information Model of Prefabricated Buildings and Its Application in Graph Database

Contact Info

Product

Resources

About