Plenary Speakers
Tina Eliassi-Rad
Northeastern University
- Title: Rethinking Optimization through Hypergraphs
- Abstract: Hypergraphs extend graphs by allowing hyperedges to connect any number of nodes, naturally capturing higher-order relationships. This talk presents three lines of work that leverage hypergraphs for optimization. First, we formulate team formation--assigning agents to tasks under energy constraints--as a constrained hypergraph discovery problem, maximizing resilience via the algebraic connectivity of the hypergraph Laplacian. A constrained simulated annealing algorithm outperforms greedy baselines on scientific collaboration datasets. Second, we show that edge-dependent vertex weight hypergraphs offer a richer representation for domains such as single-cell RNA sequencing, where random walks on a hypergraph (with cells as nodes and genes as hyperedges) yield cell embeddings that improve clustering compared to standard co-expression graphs. Third, we introduce HypOp, a distributed learning-based solver for constrained combinatorial optimization that models problems as constraint hypergraphs and uses hypergraph neural networks to find solutions. HypOp achieves competitive performance with significantly lower runtime than simulated annealing and gradient descent baselines, scales through federated distributed training, and supports transfer learning across different optimization problems on the same graph. Together, these results demonstrate that rethinking optimization through hypergraphs enables more expressive representations, more resilient solutions, and more scalable algorithms.
- Bio: Tina Eliassi-Rad is a Professor of Computer Science and The Inaugural Joseph E. Aoun Chair at Northeastern University. She is also a core faculty member at Northeastern's Network Science Institute. In addition, she is an external faculty member at the Santa Fe Institute and the Vermont Complex Systems Institute. Prior to joining Northeastern, Tina was an Associate Professor of Computer Science at Rutgers University; and before that she was a member of technical staff and principal investigator at Lawrence Livermore National Laboratory. Tina earned her Ph.D. in Computer Sciences (with a minor in Mathematical Statistics) at the University of Wisconsin-Madison. Her research is at the intersection of data mining, machine learning, and network science. She has over 150 peer-reviewed publications (including a few best paper and best paper runner-up awards); and has given over 300 invited talks and 14 tutorials. Tina's work has been applied to personalized search on the World-Wide Web, statistical indices of large-scale scientific simulation data, fraud detection, mobile ad targeting, cyber situational awareness, drug discovery, democracy and online discourse, and ethics in machine learning. Her algorithms have been incorporated into systems used by governments and industry (e.g., IBM System G Graph Analytics), as well as open-source software (e.g., Stanford Network Analysis Project). Tina received an Outstanding Mentor Award from the U.S. Department of Energy's Office of Science in 2010, became an ISI Foundation Fellow in 2019, was named one of the 100 Brilliant Women in AI Ethics in 2021, received Northeastern University's Excellence in Research and Creative Activity Award in 2022, was awarded the Lagrange Prize in 2023, and was elected Fellow of the Network Science Society in 2023.
- Web page: https://eliassi.org/
Przemysław Kazienko
Wroclaw University of Science and Technology
- Title: AI meets network science
- Abstract: Network Science and AI have traditionally approached complex data, but using different methodologies — the former focusing on structural properties and mathematical models, the latter on generalized function approximation and statistical inference. This paradigm is shifting with the maturation of Graph Representation Learning, in which network components (nodes, edges, or entire graphs) are transformed into a continuous vector space. This approach makes it much easier to extend these components with any attributes (e.g. weighted graphs). Learning representations also poses several problems for dynamic structures, such as temporal graphs, which require regularized alignment. Simultaneously, there are two relatively new but interlinked domains in AI: foundation models and generative models (GenAI). By applying these concepts to graph structures, we obtain Graph Foundation Models (GFMs) that provide scalable, general-purpose intelligence for structured data, enabling broad transfer across graph-centric tasks and domains. Foundation models are pre-trained on a large number of graphs (or subgraphs), so they are capable of capturing general knowledge across domains, which is then exploited for simple downstream tasks or generative tasks. Appropriate learning and inference stages will be enumerated. GFM models can also be integrated with natural language, making it easier to generate or even improve graph structures with natural-language prompts. The lecture will also cover some additional topics related to rational and non-rational learning, cognitive science, and the contribution of network science to the problem of AI model structures.
- Bio: Przemysław (Przemek) Kazienko, Ph.D. is a full professor and leader of three research groups: Impact AI (AI impact on humans, social influence), HumaNLP (human-centred NLP, LLMs), and Emognition (affective computing, fundamental models for physiological signals) at Wroclaw Tech (Wroclaw University of Science and Technology), Poland. He has authored over 300 research papers, including 60+ in journals with impact factor related to social/complex network analysis, complex networks, personalization and subjective tasks in NLP, Large Language Models (LLMs), self-learning LLMs, hallucination, ethics and responsibility in AI, affective computing and emotion recognition, deep machine learning, sentiment analysis, collaborative systems, recommender systems, information retrieval, data security, and many others. He delivered over 40 keynote and invited talks to international audiences and served as co-chair for more than 20 international scientific conferences and workshops. He initiated and led over 50 projects, including large European ones, primarily in collaboration with companies with a total local budget exceeding €10M. He is an IEEE Senior Member, a member of the Polish Committee for Standardization in AI, and the Ethics Committee for the LLM development. He has been on the board of Network Science Society for several years.
- Web page: https://kazienko.eu/en
Renaud Lambiotte
University of Oxford
- Title: From Signs to Matrices on Edges: Generalising Structural Balance on Networks
- Abstract: Structural balance is a classical notion in network science, originally introduced in social psychology to describe globally consistent patterns of positive and negative relations. Here we present an overview of a series of works that progressively generalise this concept from signed graphs to richer classes of weighted networks. Starting from weighted signed networks, where each edge carries a real value with a sign, we revisit the classification into balanced, antibalanced, and strictly unbalanced regimes, and show how each is reflected in spectral properties and in the behaviour of spreading processes and other dynamics. We then extend the framework to networks whose edges are weighted by complex numbers, replacing the binary sign with a continuous phase, and further to matrix-weighted networks, where interactions act on multidimensional states through matrix couplings. In all these settings, the same fundamental principle applies: balance, or more generally coherence, describes whether signals propagating along different paths combine coherently or destructively interfere. This unifying viewpoint connects structural consistency to the spectrum of generalised adjacency and Laplacian matrices, and thereby to the long-term behaviour of both linear and nonlinear dynamics. We illustrate the reach of this perspective through consensus dynamics, random walks, spectral clustering, and synchronisation of higher-dimensional Kuramoto oscillators on networks.
- Bio: Renaud Lambiotte has a PhD in Physics from the Université Libre de Bruxelles. Following postdocs at ENS Lyon, Université de Liège, UCLouvain and Imperial College London, and a Professorship in Mathematics at the University of Namur, he is currently Professor of Networks and Nonlinear Systems at the Mathematical Institute of Oxford University. His main research interests are the modelling and analysis of large networks, with a particular focus on clustering and temporal networks, and applications in social and neuronal systems. He is Associate Editor for Science Advances, an INET Fellow, External Faculty at the Complexity Hub in Vienna and Teaching Fellow at Somerville College.
- Web page: https://www.maths.ox.ac.uk/people/renaud.lambiotte
Cristopher Moore
Santa Fe Institute
- Title: Which links matter most? Sparsifying epidemic models with effective resistance
- Abstract: Network science has increasingly become central to the field of epidemiology. However, many networks derived from modern datasets are not just large, but dense, with a high average degree. One way to reduce the computational cost of simulating epidemics on these networks is sparsification, where a subset of edges is selected and reweighted based on some measure of their importance. Following recent work in computer science, we find that the most accurate approach uses the effective resistances of edges, which can be computed from the graph Laplacian. The resulting sparse network preserves both the local and global behavior of the SIR model, including the probability each node becomes infected and its distribution of arrival times. This holds even when the sparse network preserves less than 10% of the edges of a mobility network from the United States. Our work helps illuminate which links of a network are most important to disease spread. Defining edge importance using purely topological methods, or by thresholding edge weights, does not perform nearly as well. I will end by discussing the possibility of using sparsification to “denoise” networks from bioinformatics. This is joint work with Alexander Mercier (Harvard School of Public Health) and Sam Scarpino (Northeastern).
- Bio: Cristopher Moore received his B.A. in Physics, Mathematics, and Integrated Science from Northwestern University, and his Ph.D. in Physics from Cornell. From 2000 to 2012 he was a professor at the University of New Mexico, with joint appointments in Computer Science and Physics. Since 2012, Moore has been a resident professor at the Santa Fe Institute. He has also held visiting positions at the Niels Bohr Institute, École Normale Superieure, École Polytechnique, Université Paris 7, Northeastern University, the University of Michigan, and Microsoft Research. Moore has written over 170 papers at the boundary between mathematics, physics, and computer science, ranging from quantum computing, social networks, and phase transitions in NP-complete problems and Bayesian inference, to risk assessment in criminal justice. He is an elected Fellow of the American Physical Society, the American Mathematical Society, and the American Association for the Advancement of Science. With Stephan Mertens, he is the author of The Nature of Computation from Oxford University Press.
- Web page: https://sites.santafe.edu/~moore/
Clara Stegehuis
University of Twente
- Title: Homophily within and across groups
- Abstract: Traditional social network analysis often models homophily, the tendency of similar individuals to form connections. using a single parameter. We will show that in many important applications, such as hypergraphs or temporal contact networks, homophily occurs at several different scales. We present a model that integrates these different homophily values through a random graph model with a maximum entropy approach. We demonstrate that the interaction between different levels of homophily has a non-trivial effect on percolation thresholds. Furthermore, we show that our model fits remarkably well on a wide range of data sets, capturing their homophily patterns accurately.
- Bio: Clara Stegehuis is an associate professor in applied mathematics at the University of Twente. She works at the intersection of probability theory, optimization and stochastic networks. She works on random graph models for various real-world problems related to for example epidemic spreading, network motifs, chemical reactions and cellular networks. Besides her mathematical work, she has a passion for science communication. She often presents the fascinating aspects of mathematics or networks at various places, from music festivals to primary schools.
- Web page: https://www.clarastegehuis.nl/