Paper submission deadline: 20.11.2024 (AoE) 27.11.2024 (AoE)
Paper author notification: 24.01.2025
NEW: If you need a further extension to submit your work, please contact us before 27.11.2024 and you will receive additional instructions.
https://www.esann.org/author_guidelines
Both in-person and online participation will be possible.
https://www.esann.org/submit_paper
09:10
Network Science Meets AI
Organized by: Matteo Zignani, Fragkiskos D. Malliaros, Ingo Scholtes, Roberto Interdonato, Manuel Dileo
09:10
Network Science Meets AI: A Converging Frontier
Matteo Zignani, University of Milan (Italy)
Fragkiskos D. Malliaros, Université Paris-Saclay (France)
Ingo Scholtes, Julius-Maximilians-Universität Würzburg (Germany)
Roberto Interdonato, CIRAD (France)
Manuel Dileo, University of Milan (Italy)
09:30
Learning of Probability Estimates for System and Network Reliability Analysis by Means of Matrix Learning Vector Quantization
Mandy Lange-Geisler , University of Applied Sciences Mittweida (germany )
Klaus Dohmen, University of Applied Sciences Mittweida (Germany)
Thomas Villmann, Mittweida University of Applied Sciences, Saxon Institute for Computational Intelligence and Machine Learning (Germany)
09:50
Enhancing neural link predictors for temporal knowledge graphs with temporal regularisers
Manuel Dileo, University of Milan (Italy)
Pasquale Minervini, University of Edinburgh; Miniml.AI (United Kingdom)
Matteo Zignani, University of Milan (Italy)
Sabrina Gaito, University of Milan (Italy)
10:10
Network Science Meets AI - poster spotlights
Organized by: Matteo Zignani, Fragkiskos D. Malliaros, Ingo Scholtes, Roberto Interdonato, Manuel Dileo
10:10
Hyperbolic representation learning in multi-layer tissue networks
Domonkos Pogány, Budapest University of Technology and Economics, Department of Artificial Intelligence and Systems Engineering (Hungary)
Péter Antal, Budapest University of Technology and Economics, Department of Measurement and Information Systems (Hungary)
10:11
Topology-Aware Activation Functions in Neural Networks
Pavel Snopov, University of Texas Rio Grande Valley (USA)
Oleg Musin, University of Texas Rio Grande Valley (USA)
The intersection of Network Science and Artificial Intelligence (AI) is a burgeoning field that promises to revolutionize our understanding and management of complex systems. This special session, titled ”Network Science Meets AI,” aims to bring together researchers from both disciplines to explore the synergies and innovative applications that arise from their convergence. Network Science has provided a unified framework for representing complex systems through the lens of interactions among elements, often captured by networks that evolve and different types of relationships among the same set of elements. This framework has revealed universal properties such as scale-free nature, robustness, resilience, and modular structure. It has also deepened our understanding of the dynamics and controllability of complex networks, particularly in areas like contagion processes and epidemic models. On the other hand, AI is transforming various scientific disciplines by introducing new paradigms for scientific discovery. The capabilities of AI to solve highly complex problems, such as protein structure prediction, drug discovery, and content creation through generative AI, are nearing human-level performance.
This special session is aimed at contributions at the intersection between network science and AI, where AI offers new tools to tackle problems in network science, or, vice versa, network science supports the design and understanding of AI methods. There are multiple areas where a stronger interaction between network science and AI is promising: For instance, novel paradigms to describe high-order interactions through hypergraphs or simplicial complexes are closely linked to current challenges in the domain of representation learning. This also influences solutions for problems such as label, link, and network property prediction, or for combinatorial optimization problems on networks. Simulating and understanding dynamical processes in complex networks is another key challenge in network science that can benefit from AI, particularly from the capacity to handle non-linear relationships among system elements and non-linearity in the system dynamics. Being able to infer dynamical equations in a data-driven fashion is crucial for the simulation of network dynamics and node/link attribute dynamics along with generating plausible networks related to real-world systems, such as proteins, molecules, transportation systems, or even social networks. Moreover, it is equally important to consider the contribution of network science to advance methods in AI and machine learning. The development of graph neural networks that are based on neural message passing is an important research theme in the deep learning community. Existing insights into the topology of complex networks can greatly assist the design of such deep neural networks for graph-structured data. Moreover, recent works have shown that objective functions for community detection originally devised in the network science community can be repurposed for cluster detection based on deep graph learning. Finally, deep neural networks themselves are becoming instances of complex systems, thus turning into a subject of interest in network science. To this end, network science methods can be used to understand or optimize the structure of deep neural networks. Furthermore, methods for the analysis of dynamics processes can advance the comprehension of the learning processes of the parameters as well as the dynamics of the inference phases.
Relevant topics include, but are not limited to:
Optimizing GNN architectures using network science principles
Novel message passing schemes inspired by complex network dynamics
Topology-aware neural network design
Representation learning for hypergraphs and simplicial complexes
Deep learning models for higher-order interactions
AI-driven prediction of network evolution
Deep learning approaches for temporal network embedding
Neural ODEs for network dynamics
Explainable graph machine learning algorithms
Explainability techniques for graph machine learning
Network Generation and Synthesis
Topological analysis of deep neural networks
Transfer of network science objectives to deep learning (e.g. community detection, influence maximization, epidemic modeling)
Theoretical foundations of graph machine learning methods
Causal discovery in complex networks
Topology-aware optimization algorithms
NEW: The authors of papers presented at this Special Session will be invited to submit extended versions of their works for possible inclusion in a special issue of Applied Network Science (published by Springer).
Selected papers from the ESANN 2025 will be published in a special issue of the Neurocomputing journal.
University of Milan
Université Paris-Saclay
Julius-Maximilians-Universität Würzburg
CIRAD
University of Milan