1. Node classification: one of the most popular benchmark tasks for graph neural networks. The standard way to apply graph neural networks to such a node classification task is to train graph neural networks in a fully-supervised manner, where we define the loss using a softmax classification function and negative log-likelihood loss.

Graph Neural Networks Node classification

2. Graph classification: Similar to node classification, applications on graph-level classification are popular as benchmark tasks. Historically, kernel methods were popular for graph classification, and—as a result—some of the most popular early benchmarks for graph classification were adapted from the kernel literature, such as tasks involving the classification of enzyme properties based on graph-based representations.

Graph neural networks graph classification

3. Relation prediction: graph neural networks are used in relation prediction tasks, such as recommender systems and knowledge graph completion.
Pre-training graph neural networks with a neighborhood reconstruction losses could be a useful strategy to improve performance on a downstream classification task.

Graph neural networks relation prediction

For most current applications, graph neural networks are used in one of three tasks: node classification, graph classification, or relation prediction.

University of Michigan - Ann Arbor

Graph neural networks capture the internal graph structures of natural language, like syntactic and semantic parse trees.

Graph Neural Networks

Applications and Loss Functions

It is important to research on how we can make graph neural networks more powerful than WL algorithm because there are some research that indicates that models such as message-passing graph neural networks are no more powerful than the WL algorithm.

Learn Before

Related

Learn After