Program At A Glance

Below is the conference schedule at a glance. Details about the plenary talks, oral sessions, and poster session are provided after the overview.

General Information

Lectures in the oral sessions are 20 minutes long, including Q&As. We recommend authors to aim for a 15-minute presentation leaving a few minutos for questions and switching the speaker. Posters should use A0 size in portrait orientation.

There will be two social events during the conference:

• Welcome Reception: Monday, June 8, 20:30-22:30, Casa Suecia, Calle del Marqués de Casa Riera, 4, 28014 Madrid.
• Conference Banquet: Tuesday, June 9, 21:00-23:00, Café Comercial, Glorieta de Bilbao, 7, 28004 Madrid.

Plenary Talks

Plenary Monday 9:20 - 10:20

Paolo Di Lorenzo, Sapienza University of Rome

Title: Sheaf-theoretic Signal Processing and Learning

Abstract: Classical graph signal processing (GSP) provides a powerful framework for modeling data on networks, but it is inherently limited to homogeneous signal spaces and pairwise interactions. Many modern applications, ranging from biological and social networks to distributed AI, require handling heterogeneous data and structured relationships beyond these assumptions. In this talk, we present sheaf-theoretic signal processing as a principled extension of GSP for modeling heterogeneous signals and complex interactions. By assigning vector spaces to nodes and edges, together with linear restriction maps, cellular sheaves encode geometric, semantic, and topological structure directly on graphs. This framework generalizes key GSP tools, leading to the sheaf Laplacian and the Sheaf Fourier Transform, whose spectrum captures signal inconsistency across the network. We then address the problem of learning sheaves from data, proposing scalable methods based on total variation minimization that jointly infer graph topology and inter-node alignment via efficient edge-wise and Procrustes-type solutions. We also show that connection graphs arise as a structured class of sheaves with a highly interpretable spectral characterization. Finally, we highlight applications in semantic communications and federated representation learning, where sheaf-based models enable alignment of heterogeneous latent spaces without enforcing a shared global representation, yielding improved performance in distributed settings.

Plenary Monday 15:00 - 16:00

Antonio Ortega, University of Southern California

Title: How to design fast GFTs

Abstract: In this talk, we provide an overview of recent advances for speeding up the computation of the Graph Fourier Transform (GFT). We first describe divide-and-conquer techniques that leverage graph structure, such as graph symmetries or graph decompositions via low-rank updates. For graphs whose structure does not yield sufficient speed-up in transform computation, we describe approximation methods, including direct transform approximations (via Givens rotations) and indirect methods that exploit more favorable structures (e.g., spectral sparsification). We demonstrate the advantages of these techniques in image/video coding and graph machine learning applications.

Joint work with Samuel Fernández Menduiña, Keng-Shi Lu, Darukeesan Pakiyarajah, and Eduardo Pavez.

Plenary Tuesday 9:00 - 10:00

Gonzalo Mateos, University of Rochester

Title: Concomitant Linear DAG Estimation

Abstract: We deal with the combinatorial problem of learning directed acyclic graph (DAG) structure from observational data adhering to a linear structural equation model (SEM). Leveraging advances in differentiable, nonconvex characterizations of acyclicity, recent efforts have advocated a continuous constrained optimization paradigm to efficiently explore the space of DAGs. Most existing methods employ lasso-type score functions to guide this search, which (i) require expensive penalty parameter retuning when the SEM noise variances change across problem instances; and (ii) implicitly rely on limiting homoscedasticity assumptions. In this talk, I will propose a new convex score function for sparsity-aware learning of linear DAGs, which incorporates concomitant estimation of scale and thus effectively decouples the sparsity parameter from noise levels. Regularization via a smooth, nonconvex acyclicity penalty term yields CoLiDE (Concomitant Linear DAG Estimation), a regression-based criterion amenable to efficient gradient computation and closed-form estimation of exogenous noise levels in heteroscedastic scenarios. The algorithm outperforms state-of-the-art methods without incurring added complexity, especially when the DAGs are larger, and the noise level profile is heterogeneous. CoLiDE exhibits enhanced stability manifested via reduced standard deviations in several domain-specific metrics, underscoring the robustness of the novel linear DAG estimator.

Plenary Tuesday 15:00 - 16:00

Daniel P. Palomar, The Hong Kong University of Science and Technology

Title: Graphs in Financial Markets

Abstract: Financial markets generate high-dimensional, non-Gaussian, and time-varying data that challenge classical statistical models. Graph-based representations offer a principled way to capture the dependency structure among assets. This talk surveys recent advances in learning graphs from financial data, with emphasis on three settings: (i) the Polynomial Graphical Lasso, which jointly estimates the precision matrix and graph topology by exploiting graph stationarity; (ii) heavy-tailed and structured graph learning, where a Student-t model and spectral Laplacian constraints yield robust k-component and bipartite graphs that reflect market sector structure; and (iii) time-varying graph learning, which combines a non-negative VAR(1) temporal prior with heavy-tailed likelihoods to track market dynamics, detect crises, and improve portfolio performance.

Plenary Wednesday 9:00 - 10:00

Luana Ruiz, Johns Hopkins University

Title: Distance-Preserving Graph Machine Learning

Abstract: A central challenge in graph machine learning is that standard learning-based methods capture local connectivity while distorting or ignoring the metric structure of graphs at larger scales. In this talk, I will present two lines of work that address this challenge from complementary angles. The first studies landmark-based distance-preserving embeddings on inhomogeneous random graphs, a flexible model capturing the community structure and degree variability observed in real networks. By analyzing neighborhood expansion via multi-type branching process approximations, we show that the embedding dimension required to achieve near-exact shortest-path preservation is significantly smaller than worst-case theory predicts, with the improvement governed by the graph’s connectivity structure. We further show that GNN-based approximations of landmark distances transfer effectively from small synthetic graphs to large real-world networks, offering a scalable surrogate for exact shortest-path computation. The second line of work introduces a mesoscopic graph rewiring strategy based on opinion dynamics-inspired contagion processes. By promoting node pairs with strong multi-hop reinforcement to direct neighbors, the method constructs a sparse auxiliary graph that provably improves homophily and whose edge weights reflect a bounded effective resistance. Applied to both GNNs and graph transformers, cascade rewiring yields consistent accuracy gains across benchmarks. Together, these results suggest a unified perspective on graph machine learning grounded in the preservation and exploitation of metric structure across scales.

Oral Sessions

Oral Session Mon-1 (Monday 10:20 - 11:20) - Joint Time-Vertex GSP

• Subspace Projection Methods for Fast Spectral Embeddings of Evolving Graphs.
  Mohammad Eini (Michigan State University); Abdullah Karaaslanli (Michigan State University); Vassilis Kalantzis (IBM Research); Panagiotis Traganitis (Michigan State University).
• Locally Stationary Time-Vertex Process Models.
  Deniz Aslan (Middle East Technical University); Elif Vural (Middle East Technical University).
• Conformal Inference for Graphs.
  Sundeep Prabhakar Chepuri (Indian Institute of Science, Bangalore); Sonakshi Dua (Indian Institute of Science, Bangalore); Gonzalo Mateos (University of Rochester).

Oral Session Mon-2 (Monday 11:50 - 13:30) - GSP Theory

• Möbius Model for Graph Signal Processing on Weighted DAGs.
  Vedran Mihal (ETH Zurich); Markus Püschel (ETH Zurich).
• Graph-Aware Diffusion for Signal Generation.
  Vimal Kumarasamy Balasubramanian (TU Delft); Sergio Rozada (URJC); Antonio G. Marques (URJC); Elvin Isufi (TU Delft); Hadi Jamali Rad (TU Delft); Andrea Cavallo (TU Delft).
• Learning Dirac Spectral Transforms for Topological Signals.
  Leonardo Di Nino (Sapienza, Università di Roma); Tiziana Cattai (Sapienza, Università di Roma); Sergio Barbarossa (Sapienza, Università di Roma); Ginestra Bianconi (School of Mathematical Sciences, Queen Mary University of London, UK); Paolo Di Lorenzo (Sapienza, Università di Roma).
• Sampling in the Graph Signal Processing Companion Model.
  John Shi (Carnegie Mellon University); Jose Moura (Carnegie Mellon University).
• Optimal Wiener-Filter Solutions for Denoising of Graph Signals on Directed Graphs.
  Chun Hei Michael Chan (EPFL); Alexandre Cionca (EPFL); Dimitri Van De Ville (EPFL).

Oral Session Mon-3 (Monday 16:00 - 17:00) - Higher Order SP & Topological SP I

• Don’t be Afraid of Cell Complexes! An Introduction to Cell Complexes and Topological Signal Processing from an Applied Perspective.
  Josef Hoppe (RWTH Aachen University); Vincent P. Grande (RWTH Aachen University); Michael T. Schaub (RWTH Aachen University).
• Scalable Higher-Order Topology Identification from Nodal Observations.
  Geert Leus (TU Delft); Elvin Isufi (TU Delft).
• Stationarity and Spectral Characterization of Random Signals on Simplicial Complexes.
  Madeline Navarro (Rice University); Andrei Buciulea Vlas (Universidad Rey Juan Carlos); Santiago Segarra (Rice University); Antonio G. Marques (Universidad Rey Juan Carlos).

Oral Session Mon-4 (Monday 17:30 - 18:30) - Higher Order SP & Topological SP II

• Vertex-frequency Hypergraph Signal Processing: Analytic Tools and Applications.
  Alcebiades Dal Col (Federal University of Espirito Santo); Fabiano Petronetto (Federal University of Espirito Santo); José R. de Oliveira Neto (Federal University of Pernambuco); Juliano B. Lima (Federal University of Pernambuco).
• Framework for Directed Hypergraph Signal Processing via tensor t-SVD.
  Carlos Mundo Levano (University of Delaware); Nicolas Bello (University of Delaware); Dan Lau (University of Kentucky); Gonzalo Arce (University of Delaware).
• Processing Probabilistic Signals on Causal Abstraction Networks.
  Gabriele D’Acunto (Sapienza University); Paolo Di Lorenzo (Sapienza University); Sergio Barbarossa (Sapienza University).

Oral Session Tue-1 (Tuesday 10:00 - 11:00) - Non-Linear GSP

• Nonstationary Graph Filters Based on Localized Frames.
  Philipp Reingruber (TU Wien); Gerald Matz (TU Wien).
• Covariance Scattering Transforms.
  Andrea Cavallo (Delft University of Technology); Elvin Isufi (Delft University of Technology).
• Sample entropy for graph signals: An approach to nonlinear analysis of graph signals.
  Mei san lei (University of Edinburgh); John Stewart Fabila-Carrasco (University of Cardiff); Javier Escudero (University of Edinburgh).

Oral Session Tue-2 (Tuesday 16:00 - 17:00) - Geometric Deep Learning I

• Unrolling Dynamic Programming via Graph Filters.
  Sergio Rozada (URJC); Samuel Rey (URJC); Gonzalo Mateos (University of Rochester); Antonio G. Marques (URJC).
• Advection–Diffusion on Graphs: A Bakry–Émery Laplacian for Spectral Graph Neural Networks.
  Mia Zosso (EPFL); Pierre Vandergheynst (EPFL); Victor Kawasaki-Borruat (EPFL); Ali Hariri (EPFL); Pierre-Gabriel Berlureau (ENS).
• Adaptive Node Feature Selection for Graph Neural Networks.
  Madeline Navarro (Rice University).

Oral Session Tue-3 (Tuesday 17:30 - 18:30) - Higher Order SP & Topological SP III

• Joint Simplicial Complex Learning via Binary Linear Programming.
  Varun Sarathchandran (Delft University of Technology); Geert Leus (Delft University of Technology).
• Cross-Laplacians Based Topological Signal Processing over Cell MultiComplexes.
  Stefania Sardellitti (University Mercatorum); Breno C. Bispo (Federal University of Pernambuco); Fernando A. N. Santos (University of Amsterdam); Juliano B. Lima (Federal University of Pernambuco).
• Hodge-Aware Surrogates for Testing Stationarity in Topological Signals.
  Flavia Petruso (EPFL); Chun Hei Michael Chan (EPFL); Dimitri Van De Ville (EPFL).

Oral Session Wed-1 (Wednesday 10:00 - 11:00) - Graph Learning

• A Covariance Matching Approach to Graph Topology Identification.
  Yongsheng Han (TU Delft); Geert Leus (TU Delft); Raj Rajan (TU Delft).
• BUILD with precision: Bottom-up inference of linear DAGs.
  Hamed Ajorlou (University of Rochester); Samuel Rey (King Juan Carlos University); Gonzalo Mateos (University of Rochester); Geert Leus (TU Delft); Antonio G. Marques (King Juan Carlos University).
• Sparsity-Aware Extended Kalman Filter for Tracking Dynamic Graphs.
  Lital Dabush (Ben-Gurion University); Nir Shlezinger (Ben-Gurion University); Tirza Routtenberg (Ben-Gurion University).

Oral Session Wed-2 (Wednesday 11:30 - 13:30) - Geometric Deep Learning II

• Fixed Aggregation Features Can Rival GNNs.
  Celia Rubio-Madrigal (CISPA Helmholtz Center for Information Security); Rebekka Burkholz (CISPA Helmholtz Center for Information Security).
• Size Transferability of Graph Transformers with Convolutional Positional Encodings.
  Javier Porras Valenzuela (University of Pennsylvania); Zhiyang Wang (University of California San Diego); Xiaotao Shang (University of Pennsylvania); Yusu Wang (University of California San Diego); Alejandro Ribeiro (University of Pennsylvania).
• A Graph Attention Network Approach to Super-Resolution Spatial Transcriptomic Data.
  Luis Alonso (University of Navarra); Mikel Hernaez (University of Navarra); Idoia Ochoa (University of Navarra).
• L2G-Net: Local to Global Spectral Graph Neural Networks via Cauchy Factorizations.
  Samuel Fernandez (University of Southern California); Eduardo Pavez (University of Southern California); Antonio Ortega (University of Southern California).
• Graph Signal Diffusion Models for Wireless Resource Allocation.
  Yigit Berkay Uslu (University of Pennsylvania); Samar Hadou (University of Pennsylvania); Shirin Saeedi Bidokhti (University of Pennsylvania); Alejandro Ribeiro (University of Pennsylvania).
• On the Effectiveness of Pretraining for Graph Combinatorial Optimization.
  David Aguado (Universidad Politécnica de Madrid); Daniel Fuertes (Universidad Politécnica de Madrid); Carlos R. del-Blanco (Universidad Politécnica de Madrid); Fernando Jaureguizar (Universidad Politécnica de Madrid).

Oral Session Wed-3 (Wednesday 15:00 - 16:00) - Higher Order SP & Topological SP IV

• STORM: Simplicial Topological Recurrent Model for Dynamics on Higher-Order Domains.
  Mohamed Salah Jebali (TU Delft); Elvin Isufi (TU Delft); Claudio Battiloro (Harvard).
• A Graph-Structured VAR Model for Data with Higher Order Temporal Dependencies and Heavy Tails.
  Amirhossein Javaheri (KTH Royal Institute of Technology); Saikat Chatterjee (KTH Royal Institute of Technology); Daniel Palomar (Hong Kong University of Science and Technology).
• A Framework for Directed Acyclic Hypergraph Learning.
  Zhiyuan Dong (University of Delaware); Carlos Mundo-Levano (University of Delaware); Gonzalo R. Arce (University of Delaware); Wei Qian (University of Delaware); Daniel Lau (University of Kentucky).

Poster Session

Poster Session (Tuesday 11:30 - 13:30)

• Topological Kalman Filtering on Cell Complexes.
  Chengen Liu (Delft University of Technology).
• Learning Graph Topology with Functional Priors: A Graph Formation Model Perspective.
  Chenyue Zhang (Chinese University of Hong Kong); Shangyuan Liu (Chinese University of Hong Kong); Hoi-To Wai (Chinese University of Hong Kong); Anthony Man-Cho So (Chinese University of Hong Kong).
• GNNs Getting ComFy: Community and Feature Similarity Guided Rewiring.
  Celia Rubio-Madrigal (CISPA Helmholtz Center for Information Security); Adarsh Jamadandi (IRISA, University of Rennes); Rebekka Burkholz (CISPA Helmholtz Center for Information Security).
• Distance-Misaligned Training in Graph Transformers and Adaptive Graph-Aware Control.
  Qinhan Hou (University of Helsinki); Jing Tang (University of Helsinki).
• Data-Driven Higher-Order Topology Learning for Leak Detection in Dynamic Water Distribution Networks.
  Tiziana Cattai (Sapienza University of Rome); Stefania Sardellitti (Universitas Mercatorum); Stefania Colonnese (Sapienza University of Rome); Francesca Cuomo (Sapienza University of Rome); Sergio Barbarossa (Sapienza University of Rome).
• ADAPTIVEMIXGNN: Local Adaptive Inductive Bias for Heterophilic Node Classification.
  Miguel Alcocer (King Juan Carlos University, Madrid); Javier Muñoz (King Juan Carlos University, Madrid); Álvaro Morán (King Juan Carlos University, Madrid).
• Planar Horizontal Visibility Graphs for Chromatin Dynamics Analysis: Applications to Cellular Metabolic States.
  Lucía Benito (Universidad Francisco de Vitoria); Diego Herráez (Universidad Francisco de Vitoria).
• A Sheaf-Theoretic Framework for Distributed Multi-Site Channel Charting.
  Enrico Grimaldi (Sapienza University of Rome); Leonardo Di Nino (Sapienza University of Rome); Mario Edoardo Pandolfo (Sapienza University of Rome); Gabriele D’Acunto (Sapienza University of Rome); Sergio Barbarossa (Sapienza University of Rome); Paolo Di Lorenzo (Sapienza University of Rome).
• Shattering the Speed-Accuracy Dichotomy in Asymmetric Routing via Anisotropic GNNs.
  Gonzalo Mantiñán Suárez (Universidad Politécnica de Madrid); Daniel Fuertes (Universidad Politécnica de Madrid); Carlos R. del-Blanco (Universidad Politécnica de Madrid); Fernando Jaureguizar (Universidad Politécnica de Madrid).
• Precision Neural Networks: Joint Graph and Relational Learning.
  Andrea Cavallo (Delft University of Technology).
• Learning Dynamics in Streaming Weighted Higher-Order Networks.
  Rohan Thekkemarickal Money (Simula); Baltasar Beferull-Lozano (Simula Metropolitan Center for Digital Engineering); Elvin Isufi (TU Delft).
• Enhancing Transformer-based Routing by Encoding Distance via Relative Positional Encoding.
  Leyre Encío (Universidad Politécnica de Madrid); Daniel Fuertes (Universidad Politécnica de Madrid); Carlos R. del-Blanco (Universidad Politécnica de Madrid); Fernando Jaureguizar (Universidad Politécnica de Madrid).
• Random Spectral Features for Graph Kernel Machines.
  Valentin de Bassompierre (UCLouvain); Laurent Jacques (UCLouvain); Jean-Charles Delvenne (UCLouvain).