Publication: Deep Learning for Dynamic Graphs: Models and Benchmarks

03 April 2024

Graphs are powerful tools to represent systems of relations and interactions, across several application fields where deep learning for graphs has found successful application, such as biology, social science and human mobility.

The key challenge when learning from graph data is how to numerically represent the combinatorial structures for effective processing and prediction by the model. A classical predictive task of molecule solubility prediction, for instance, requires the model to encode both topological information and chemical properties of atoms and bonds. Graph representation learning solves the problem in a data-driven fashion, by learning a mapping function that compresses the complex relational information of a graph into an information-rich feature vector that reflects both structural and label information in the original graph.

Recent progress in research on deep graph networks (DGNs) has led to a maturation of the domain of learning on graphs. Despite the growth of this research field, there are still important challenges that are yet unsolved. Specifically, there is an urge of making DGNs suitable for predictive tasks on real-world systems of interconnected entities, which evolve over time.

With the aim of fostering research in the domain of dynamic graphs, EMERGE partners from the University of Pisa survey recent advantages in learning both temporal and spatial information, providing a comprehensive overview of the current state-of-the-art in the domain of representation learning for dynamic graphs. The authors also conduct a fair performance comparison among the most popular proposed approaches on node-and edge-level tasks, leveraging rigorous model selection and assessment for all the methods, thus establishing a sound baseline for evaluating new architectures and approaches.

Read the paper in the link below.