June 30^th, 2025

Venue: Escola de Enxeñaría de Telecomunicación, Salón de Grados and Sala de Reuniones.

• 16:00 – 19:00 Red COMONSENS Networking sessions.
• 17:00 – 19:00 Red COMONSENS Steering meeting.

July 1^st, 2025

Venue: Sede Afundación Vigo. Talks at Sala de Conferencias. Poster session at the Garden.

• 08:30 – 9:00 Registration and Welcome.
• 9:00 – 10:00 Talk by Robert Heath (University of California, San Diego, USA).
  • Title: Embracing Reconfigurable Antennas in the Tri-hybrid MIMO Architecture for 6G
  • Abstract: Multiple-input multiple-output (MIMO) communication has driven major advancements in data rates and spectrum efficiency. As MIMO systems scale to support more antennas and higher frequencies, though, power consumption and hardware complexity become bottlenecks. In this talk, I will unveil the tri-hybrid MIMO architecture that integrates low-power reconfigurable antennas with both digital and analog precoding. First, I will provide some background on the fundamentals of reconfigurable antennas. Second, I will introduce the tri-hybrid architecture, highlighting the benefits in terms of power and aperture. Third, I will describe an initial version of the tri-hybrid MIMO architecture that makes use of dynamic metasurface antennas. This preliminary architecture serves as a case study to illustrate key performance. Finally, I will discuss future challenges in designing reconfigurable arrays and optimizing tunable antenna parameters. The tri-hybrid architecture introduces new research directions in physically-consistent wireless communications at the intersection of electromagnetics, circuits, and signal processing.
• 10:15 – 11:15 Talk by Jana de Wiljes (Ilmenau University of Technology, Germany).
  • Title: Long-Time Stability and Accuracy of Gaussian Filters for Randomised Observations
  • Abstract: The seamless integration of large datasets into computational models is a critical task across a broad range of applications. One of the main challenges lies in handling the complexity of high-dimensional nonlinear filtering for sophisticated and computationally intensive evolution models. Despite the frequent violation of theoretical assumptions in practical settings, Gaussian approximative filters are widely regarded as state-of-the-art. Their effectiveness in highly nonlinear contexts with large state spaces highlights their practical value. Recent research has demonstrated that these filters perform well in terms of tracking accuracy for nonlinear evolution models. In this talk, we will present a key result: specific bounds tailored to different filter variants that confirm their long-time stability and accuracy. While the robustness of Gaussian approximative filters is well established, there is a growing interest in developing filters that offer improved accuracy without sacrificing stability. We will explore and present an adaptive approach using partial and randomised observations to achieve this goal.
• 11:15 – 11:45 Coffee break.
• 11:45 – 12:45 Talk by Daniel Romero (University of Agder, Norway).
  • Title: Transformer Networks for Spatial Field Estimation
  • Abstract: The recent popularity of transformer deep neural networks has been propelled by the unprecedented success of chatbots such as ChatGPT, which are built upon this kind of architecture. However, beyond large language models, the idea underlying transformers can be extended to other problems. In our work, transformers are applied to the estimation of spatial fields given noisy observations at a finite set of locations. Unlike conventional deep architectures, the proposed algorithm does not require spatial discretization, can accommodate an arbitrary number of observations, and does not involve backward passes for context adaptation. Rotation and translation equivariance are achieved by means of simple feature transformations. With the K-nearest neighbors algorithm as a departure point, the first part of this talk provides a tutorial introduction to transformers that gradually builds attention-based networks via a sequence of small, intuitive steps. The second part details how transformers can be used for spatial field estimation and active sensing. Special focus is placed on the radio map estimation (RME) problem, where the proposed estimator sets the state of the art.
• 13:00 – 15:00 Free time.
• 15:00 – 17:00 Poster session.
• 21:00 Social dinner: O Rei Pescador, Praza Compostela 29, 36201 Vigo
  

July 2^nd, 2025

Venue: Sede Afundación Vigo. Seminar at Sala de Conferencias.

• 9:00 – 11:30 Talk/Seminar by Gonzalo Mateos (University of Rochester, USA).
  • Title: Graph adjacency spectral embeddings: Algorithmic advances and applications
  • Abstract: The random dot product graph (RDPG) is a tractable yet expressive generative graph model for relational data, that subsumes simple Erdős-Rényi and stochastic block model ensembles as particular cases. RDPGs postulate that there exist latent positions for each node and specify the edge formation probabilities via the inner product of the corresponding latent vectors. In this tutorial, we first focus on the embedding task of estimating these latent positions from observed graphs. The workhorse adjacency spectral embedding (ASE) offers an approximate solution obtained via the eigendecomposition of the adjacency matrix, which enjoys solid statistical guarantees but can be computationally intensive and is formally solving a surrogate problem. To address these challenges, we bring to bear recent non-convex optimization advances and demonstrate their impact to RDPG inference. We show the proposed algorithms are scalable, robust to missing network data, and can track the latent positions over time when the graphs are acquired in a streaming fashion; even for dynamic networks subject to node additions and deletions. We also discuss extensions to the vanilla RDPG to accommodate directed and weighted graphs. Unlike previous proposals, our non-parametric RDPG model for weighted networks does not require a priori specification of the weights’ distribution to perform inference and estimation in a provably consistent fashion. Finally, we discuss the problem of online monitoring and detection of changes in the underlying data distribution of a graph sequence. Our idea is to endow sequential change-point detection (CPD) techniques with a graph representation learning substrate based on the versatile RDPG model. We share an open-source implementation of the proposed node embedding and online CPD algorithms, whose effectiveness is demonstrated via synthetic and real network data experiments.
• 11:30 – 12:00 Coffee break
• 12:00 – 14:30 Tutorial by Nuria González-Prelcic (University of California, San Diego, USA).
  • Title: The Integrated Sensing and Communication Revolution for 6G and Beyond: Fundamentals and Emerging Technologies
  • Abstract: In this tutorial, I describe different frameworks for integrating sensing and communications (ISAC) in future generation wireless systems, discussing the different features to be exploited at different frequency bands. We focus on a communications-centric perspective for ISAC with tight integration of waveform as well as time and frequency resources for sensing and communications, versus other approaches where integration only appears at the site or at the spectrum level. We present first an overview of the fundamental problems and techniques for different ISAC configurations -including localization, monostatic, bistatic and multistatic sensing- and then we introduce recent advances and their limitations, which motivates the last section of the tutorial on emerging strategies for integration.
• 14:30 SIC’25 Closing

July 3^rd, 2025

Venue: Escola de Enxeñaría de Telecomunicación, Salón de Grados.

• 10:00 – 19:00 Red COMONSENS Networking sessions.

July 4^th, 2025

Venue: Escola de Enxeñaría de Telecomunicación, Salón de Grados.

• 10:00 – 15:00 Red COMONSENS Networking sessions.