Design and Theory of Deep Graph Learning
Special Session @ IEEE IJCNN 2025
Special Session @ IEEE IJCNN 2025
Design and Theory of Deep Graph Learning is a special session at the 2025 IEEE International Joint Conference on Neural Networks (IJCNN), which will be held in Rome, Italy, June 30 - July 5, 2025.
Paper Submissions: January 15th, 2025 January 30th, 2025
Paper Acceptance Notifications: March 31, 2025
Conference: June 30 - July 5, 2024
Graph structures allow to represent structured/complex data with entities and their relationships. These data structures naturally characterize a wide range of problems, including the areas of bio/chemistry (e.g. (macro-)molecules, proteins, biological networks), natural language processing (e.g. strings and parse trees), social network analysis (e.g. link prediction, object identification), information dissemination on social networks (temporal/dynamics graphs), and epidemiological studies (graph agent-based models).
In general, graphs offer an extremely flexible tool to describe directly and effectively the relationships between data items that can be lost by the ”flat” (vector) representations used by traditional Machine Learning (ML) and Neural Networks (NNs) tools. The extension of ML/NNs to graph domains makes it possible to address, in a systematic and general way, this variety of problems with data-driven approaches. By being able to deal with the inherent nature of structured data, learning models are endowed with a formidable capability and flexibility to address new domains and to improve accuracy and efficiency in solving complex problems. However, there are still many open points and challenges for basic and applied research, as the research in ML/NNs for graphs generalizes in a non-obvious way theoretical and modeling issues already mature in vector domains.
This special session at IJCNN 2025 aims to bring together cutting-edge research and new ideas in deep learning for graphs, addressing open challenges and advancing both theoretical and practical perspectives. Despite rapid progress, this field continues to present unique challenges—such as model generalization, structure learning, and the adaptation of methods developed for vector spaces to graph domains.
This session seeks to serve as a key gathering for researchers to discuss both foundational developments and innovative applications within the expanding domain of graph-based deep learning. We invite papers that present advances in model design, structure learning, graph kernels, and graph stream processing, alongside theoretical insights, benchmarks, and impactful applications across disciplines. This session offers an exceptional opportunity for researchers to share insights, highlight emerging applications, and set new directions for the future of machine learning on graphs. Topics of interest to this session include, but are not limited to:
Graph representation learning
Graph generation (probabilistic models, variational autoencoders, adversarial learning, normalizing flow, diffusion models, etc.)
Graph learning and relational inference
Graph coarsening and pooling in graph neural networks
Graph kernels and distances
Scalability, data efficiency, and training techniques of graph neural networks
Tensor methods for structured data
Theory of graph neural networks (e.g., expressive power, learnability, generalization capability, transferability, negative results)
Trustworthy methods and analysis for deep graph learning, including approaches towards explainability (XAI), accountability, and robustness
Over-smoothing and over-squashing in deep graph neural networks
Heterophilic Graph Learning: Benchmarks, Models, Theoretical Analysis, Applications and Challenges
Deep learning for dynamic graphs and spatio-temporal data
Anomaly and change detection in graph data
Reservoir computing and randomized neural networks for graphs
Recurrent, recursive, and contextual models
Graph datasets and benchmarks
Applications in natural language processing, neuroscience, computer vision (e.g. point clouds), materials science, cheminformatics, computational biology, social networks, etc.
Go to the IEEE IJCNN 2025 website and click on "Submit your paper".
You will be redirected to the CMT submission system. Log in with your account (or register in case you don't have one yet).
Select "Main Track" among the IJCNN 2025 new submission types.
Insert the details of your paper, and select "Design and Theory of Deep Graph Learning" from the Special Sessions as the principal topic.
Click on "Register Paper". Good Luck!!
Ming Li (Zhejiang Normal University)
Pietro Liò (University of Cambridge)
Alessio Micheli (University of Pisa)
Nicolò Navarin (University of Padua)
Luca Pasa (University of Padua)
Davide Rigoni (University of Padua)
Franco Scarselli (University of Siena)
Alessandro Sperduti (University of Padua)
Domenico Tortorella (University of Pisa)
For any enquire, please write to domenico [dot] tortorella [at] phd [dot] unipi [dot] it or davide [dot] rigoni [dot] 1 [at] unipd [dot] it