Megnn: Meta-path extracted graph neural network for heterogeneous graph representation learning

Chang, Yaomin; Chen, Chuan; Hu, Weibo; Zheng, Zibin; Zhou, Xiaocong; Chen, Shouzhi

doi:10.1016/j.knosys.2021.107611

Cited by 55 publications

(17 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using the message passing paradigm of GNN through trainable convolutions, Megnn can optimize and extract effective metapaths for heterogeneous graph representation learning. A large number of experimental results on three datasets not only demonstrate the effectiveness of Megnn's method compared with the latest method but also prove that the extracted metapath has good interpretability [9]. Lei et al used source task models to help reduce the training cost of target task models in order to reduce the neural network training cost.…”

Section: Related Workmentioning

confidence: 97%

Classification Algorithm for Heterogeneous Network Data Streams Based on Big Data Active Learning

Zhan

2022

Journal of Applied Mathematics

View full text Add to dashboard Cite

Data classification is one of the main tasks in the current data mining field, and the existing network data triage algorithms have problems such as too small a proportion of labeled samples, a large amount of noise, and redundant data, which lead to low classification accuracy of data stream implementation. Network embedding can effectively improve these problems, but the network embedding itself has problems such as capturing relational honor and ambiguity. This study proposes a SNN-RODE based LapRLS heterogeneous network data classification algorithm to achieve deep embedding of structure and semantics among nodes by constructing a multitask SNN and selecting dead song datasets to perform mining tasks to train the neural network. Then a semisupervised learning classifier based on Laplace regular least squares regression model is designed to use the relative support difference function as the decision method and optimize the function. The simulation experimental results show that the SNN-RODE-LapRLS algorithm improves the performance by 14%-51% over the mainstream classification algorithms, and the consumption time meets the demand of real-time classification.

show abstract

Section: Related Workmentioning

confidence: 97%

Classification Algorithm for Heterogeneous Network Data Streams Based on Big Data Active Learning

Zhan

2022

Journal of Applied Mathematics

View full text Add to dashboard Cite

show abstract

“…Large-scale heterogeneous graph neural networks. Many heterogeneous GNN structures are proposed in recent years [16,39,106,126,137]. However, few distributed systems take the unique characteristic of heterogeneous graphs into consideration to support heterogeneous GNNs.…”

Section: Future Directionmentioning

confidence: 99%

Distributed Graph Neural Network Training: A Survey

Shao¹,

Li²,

Gu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Graph neural networks (GNNs) are a type of deep learning models that learning over graphs, and have been successfully applied in many domains. Despite the effectiveness of GNNs, it is still challenging for GNNs to efficiently scale to large graphs. As a remedy, distributed computing becomes a promising solution of training large-scale GNNs, since it is able to provide abundant computing resources. However, the dependency of graph structure increases the difficulty of achieving high-efficiency distributed GNN training, which suffers from the massive communication and workload imbalance. In recent years, many efforts have been made on distributed GNN training, and an array of training algorithms and systems have been proposed. Yet, there is a lack of systematic review on the optimization techniques from graph processing to distributed execution. In this survey, we analyze three major challenges in distributed GNN training that are massive feature communication, the loss of model accuracy and workload imbalance. Then we introduce a new taxonomy for the optimization techniques in distributed GNN training that address the above challenges. The new taxonomy classifies existing techniques into four categories that are GNN data partition, GNN batch generation, GNN execution model, and GNN communication protocol. We carefully discuss the techniques in each category. In the end, we summarize existing distributed GNN systems for multi-GPUs, GPU-clusters and CPU-clusters, respectively, and give a discussion about the future direction on scalable GNNs.

show abstract

“…Since GCA [56] is for node classification, we extend it to the subgraph classification task by adding an average pooling layer. For SubGNN [1], we use the optimal model hyperparameters suggested in the official source code 5 .…”

Section: Experimental Settingsmentioning

confidence: 99%

“…Powerful GRLs are usually built on large amounts of training data with supervised signals (labels), but real-world applications face the problem of lacking labeled data. Recent studies notice this challenge in GRL and have proposed some approaches for data-efficient learning on graphs (GEL) [5,18,41,52,53]. However, existing GEL methods are node-level [54], edge-level [17], or graph-level [32].…”

Section: Introductionmentioning

confidence: 99%

Position-Aware Subgraph Neural Networks with Data-Efficient Learning

Liu¹,

Yang²,

Xie³

et al. 2022

Preprint

View full text Add to dashboard Cite

Data-efficient learning on graphs (GEL) is essential in real-world applications. Existing GEL methods focus on learning useful representations for nodes, edges, or entire graphs with "small" labeled data. But the problem of data-efficient learning for subgraph prediction has not been explored. The challenges of this problem lie in the following aspects: 1) It is crucial for subgraphs to learn positional features to acquire structural information in the base graph in which they exist. Although the existing subgraph neural network method is capable of learning disentangled position encodings, the overall computational complexity is very high. 2) Prevailing graph augmentation methods for GEL, including rule-based, sample-based, adaptive, and automated methods, are not suitable for augmenting subgraphs because a subgraph contains fewer nodes but richer information such as position, neighbor, and structure. Subgraph augmentation is more susceptible to undesirable perturbations. 3) Only a small number of nodes in the base graph are contained in subgraphs, which leads to a potential "bias" problem that the subgraph representation learning is dominated by these "hot" nodes. By contrast, the remaining nodes fail to be fully learned, which reduces the generalization ability of subgraph representation learning. In this paper, we aim to address the challenges above and propose a Position-Aware Data-Efficient Learning framework for subgraph neural networks called PADEL. Specifically, we propose a novel node position encoding method that is anchor-free, and design a new generative subgraph augmentation method based on a diffused variational subgraph autoencoder, and we propose exploratory and

show abstract

Megnn: Meta-path extracted graph neural network for heterogeneous graph representation learning

Cited by 55 publications

References 18 publications

Classification Algorithm for Heterogeneous Network Data Streams Based on Big Data Active Learning

Classification Algorithm for Heterogeneous Network Data Streams Based on Big Data Active Learning

Distributed Graph Neural Network Training: A Survey

Position-Aware Subgraph Neural Networks with Data-Efficient Learning

Contact Info

Product

Resources

About