Semi-Supervised and Self-Supervised Classification with Multi-View Graph Neural Networks

Yuan, Jinliang; Yu, Hualei; Cao, Meng; Xu, Ming; Xie, Jin; Wang, Chongjun

doi:10.1145/3459637.3482477

Cited by 11 publications

(4 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multiple datasets demonstrate that the proposed model is significantly superior to the conventional graph neural network model. Based on the remarkable success of graph neural network in processing graph structure data, Yuan et al [7] designed GNN with strong representation capability in response to the node classification problem. However, the depth model has a problem with overfitting, so a novel concept of aggregating more useful information based on multiple views without depth structure is proposed, and a large number of node classification experiments are conducted on six public datasets, demonstrating the superiority of the proposed model over the most recent methods.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Enhancement Economic System Based-Graph Neural Network in Stock Classification

Zhang

2023

IEEE Access

View full text Add to dashboard Cite

As a result of the integration of the stock industry into the entire international economic system, stock companies publish hundreds of prospectuses every second. The ability to quickly and accurately classify the companies and categories to which these data belong, as well as improve the performance of the economic system, has become the key to unlocking its corresponding value at this stage. In order to solve this problem, graph neural network techniques are used to accomplish the classification of stocks in the science and technology version, thus indirectly alleviating the enormous pressure on the economic system. In this study, to complete semi-supervised classification, we propose the PA-GCN model, which takes the lead in graph attention calculation of stock nodes and introduces the Elu activation function. Specifically, in this study, we constructed a stock dataset and implemented a dropout layer to prevent overfitting. The final results on the stock dataset and the publicly accessible Cora dataset demonstrate that this strategy may effectively improve the economic system and that the model has good performance. In the two dataset tests, classification accuracy can be attained at 81.69% and 81.2%, respectively. It also demonstrates the method's viability for classifying stock nodes.

show abstract

Section: Related Workmentioning

confidence: 99%

“…The attention coefficient is then normalized to yield the final attention coefficienta ij . As indicated by Equation (7). Finally, the features are then weighted and summed in accordance with the attention coefficient.…”

Section: ) Graph Attention Layermentioning

confidence: 99%

Enhancement Economic System Based-Graph Neural Network in Stock Classification

Zhang

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…For example, Wang et al (2020) proposed AM-GCN, which constructed two views from network structure and node features, respectively, and then used an attention mechanism to automatically learn the weights of the views. Yuan et al (2021) designed three complementary views and adopted convolutional operations to learn view embeddings. Finally, an attention mechanism was also used to fuse node representations for the classification task.…”

Section: Introductionmentioning

confidence: 99%

“…Although the above methods are effective, there is room for improvement. For instance, AM-GCN only constructed two views, and Yuan et al (2021) used convolution operations to simply aggregate node features and topological information, which limits the model to capture important information. Therefore, we constructed three views to learn the comprehensive drug interaction relationships.…”

Section: Introductionmentioning

confidence: 99%

MFDA: Multiview fusion based on dual-level attention for drug interaction prediction

et al. 2022

View full text Add to dashboard Cite

Drug-drug interaction prediction plays an important role in pharmacology and clinical applications. Most traditional methods predict drug interactions based on drug attributes or network structure. They usually have three limitations: 1) failing to integrate drug features and network structures well, resulting in less informative drug embeddings; 2) being restricted to a single view of drug interaction relationships; 3) ignoring the importance of different neighbors. To tackle these challenges, this paper proposed a multiview fusion based on dual-level attention to predict drug interactions (called MFDA). The MFDA first constructed multiple views for the drug interaction relationship, and then adopted a cross-fusion strategy to deeply fuse drug features with the drug interaction network under each view. To distinguish the importance of different neighbors and views, MFDA adopted a dual-level attention mechanism (node level and view level) to obtain the unified drug embedding for drug interaction prediction. Extensive experiments were conducted on real datasets, and the MFDA demonstrated superior performance compared to state-of-the-art baselines. In the multitask analysis of new drug reactions, MFDA obtained higher scores on multiple metrics. In addition, its prediction results corresponded to specific drug reaction events, which achieved more accurate predictions.

show abstract