Papers › Network In Graph Neural Network

Network In Graph Neural Network

23 Nov 2021arXiv:2111.11638archive 2025-07-28

Xiang Song, Runjie Ma, Jiahang Li, Muhan Zhang, David Paul Wipf

Graph Neural Networks (GNNs) have shown success in learning from graph structured data containing node/edge feature information, with application to social networks, recommendation, fraud detection and knowledge graph reasoning. In this regard, various strategies have been proposed in the past to improve the expressiveness of GNNs. For example, one straightforward option is to simply increase the parameter size by either expanding the hid-den dimension or increasing the number of GNN layers. However, wider hidden layers can easily lead to overfitting, and incrementally adding more GNN layers can potentially result in over-smoothing.In this paper, we present a model-agnostic methodology, namely Network In Graph Neural Network (NGNN ), that allows arbitrary GNN models to increase their model capacity by making the model deeper. However, instead of adding or widening GNN layers, NGNN deepens a GNN model by inserting non-linear feedforward neural network layer(s) within each GNN layer. An analysis of NGNN as applied to a GraphSage base GNN on ogbn-products data demonstrate that it can keep the model stable against either node feature or graph structure perturbations. Furthermore, wide-ranging evaluation results on both node classification and link prediction tasks show that NGNN works reliably across diverse GNN architectures.For instance, it improves the test accuracy of GraphSage on the ogbn-products by 1.6% and improves the hits@100 score of SEAL on ogbl-ppa by 7.08% and the hits@20 score of GraphSage+Edge-Attr on ogbl-ppi by 6.22%. And at the time of this submission, it achieved two first places on the OGB link prediction leaderboard.

PaperPDF

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Fraud DetectionGraph Neural NetworkLink PredictionNode Classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Link Property Prediction ogbl-citation2 NGNN + SEAL Ext. data No #7 of 23 Archive leaderboard report
Link Property Prediction ogbl-citation2 NGNN + SEAL Number of params 1134402 #7 of 23 Archive leaderboard report
Link Property Prediction ogbl-citation2 NGNN + SEAL Test MRR 0.8891 ± 0.0022 #7 of 23 Archive leaderboard report
Link Property Prediction ogbl-citation2 NGNN + SEAL Validation MRR 0.8879 ± 0.0022 #7 of 23 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GraphSAGE Ext. data No #20 of 34 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GraphSAGE Number of params 591873 #20 of 34 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GraphSAGE Test Hits@50 0.5359 ± 0.0056 #20 of 34 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GraphSAGE Validation Hits@50 0.6281 ± 0.0046 #20 of 34 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GCN Ext. data No #21 of 34 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GCN Number of params 428033 #21 of 34 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GCN Test Hits@50 0.5348 ± 0.0040 #21 of 34 Archive leaderboard report
Link Property Prediction ogbl-collab NGNN + GCN Validation Hits@50 0.6273 ± 0.0040 #21 of 34 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GraphSAGE Ext. data No #22 of 31 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GraphSAGE Number of params 1618433 #22 of 31 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GraphSAGE Test Hits@20 0.5770 ± 0.1523 #22 of 31 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GraphSAGE Validation Hits@20 0.7323 ± 0.0040 #22 of 31 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GCN Ext. data No #23 of 31 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GCN Number of params 1487361 #23 of 31 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GCN Test Hits@20 0.5483 ± 0.1581 #23 of 31 Archive leaderboard report
Link Property Prediction ogbl-ddi NGNN + GCN Validation Hits@20 0.7121 ± 0.0038 #23 of 31 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + SEAL Ext. data No #8 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + SEAL Number of params 735426 #8 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + SEAL Test Hits@100 0.5971 ± 0.0245 #8 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + SEAL Validation Hits@100 0.5995 ± 0.0205 #8 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GraphSAGE Ext. data No #18 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GraphSAGE Number of params 556033 #18 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GraphSAGE Test Hits@100 0.4005 ± 0.0138 #18 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GraphSAGE Validation Hits@100 0.4058 ± 0.0123 #18 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GCN Ext. data No #19 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GCN Number of params 410113 #19 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GCN Test Hits@100 0.3683 ± 0.0099 #19 of 26 Archive leaderboard report
Link Property Prediction ogbl-ppa NGNN + GCN Validation Hits@100 0.3834 ± 0.0082 #19 of 26 Archive leaderboard report
Node Property Prediction ogbn-proteins GAT+BOT+NGNN Ext. data No #5 of 26 Archive leaderboard report
Node Property Prediction ogbn-proteins GAT+BOT+NGNN Number of params 11740552 #5 of 26 Archive leaderboard report
Node Property Prediction ogbn-proteins GAT+BOT+NGNN Test ROC-AUC 0.8809 ± 0.0016 #5 of 26 Archive leaderboard report
Node Property Prediction ogbn-proteins GAT+BOT+NGNN Validation ROC-AUC 0.9375 ± 0.0019 #5 of 26 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Graph Neural NetworkGraphSAGE

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections