Statements (56)
| Predicate | Object |
|---|---|
| gptkbp:instance_of |
gptkb:television_channel
|
| gptkbp:bfsLayer |
4
|
| gptkbp:bfsParent |
gptkb:Thomas_Kipf
|
| gptkbp:application |
Social network analysis
Recommendation systems Knowledge graph completion Molecular property prediction |
| gptkbp:author |
gptkb:Yoshua_Bengio
Lev Barenboim Petar Veličković Wilker Aziz |
| gptkbp:benefits |
Computationally intensive
Focuses on important nodes Handles varying node degrees Requires careful tuning of hyperparameters |
| gptkbp:code |
gptkb:Cheb_Net
gptkb:Graph_SAGE GA Tv2 Graph Isomorphism Network |
| gptkbp:content_type |
Activation layer
Attention layer Feedforward layer Normalization layer |
| gptkbp:contribution |
Enhanced interpretability of node representations
Improved performance on benchmark datasets Introduced attention mechanism to GN Ns |
| gptkbp:data_type |
Undirected graphs
Directed graphs Weighted graphs Unweighted graphs |
| https://www.w3.org/2000/01/rdf-schema#label |
Graph Attention Networks
|
| gptkbp:input_output |
Node embeddings
Graph embeddings Node features Edge features |
| gptkbp:introduced |
gptkb:2018
|
| gptkbp:is_a_framework_for |
gptkb:Graphics_Processing_Unit
gptkb:Py_Torch |
| gptkbp:is_designed_for |
Graph-structured data
|
| gptkbp:key |
Attention mechanism
Node representation learning Scalability to large graphs |
| gptkbp:performance |
Accuracy
F1 score AUC-ROC |
| gptkbp:related_to |
Deep learning
Neural networks Graph Convolutional Networks |
| gptkbp:training |
Supervised learning
Semi-supervised learning Unsupervised learning |
| gptkbp:use_case |
Anomaly detection
Link prediction Node classification Community detection Graph classification |