Abstract
Reconfigurable Intelligent Surfaces (RIS) are an enabling technology widely investigated towards 6G. The viability of large active metasurfaces is constrained by the RF performance, cost, and power consumption. The number of switches per unit cell is a key design parameter that designers aim to minimize following cost and power consumption drivers. However, an efficient use of the aperture is ultimately required and although a one-to-one correspondence between number of switches and phase-quantization bits seems intuitive, one may question its impact. Here we present a full-wave evaluation of a 30x30 1-bit reflective RIS, implemented considering two pin diodes per unit cell. The RIS allows scanning up to 60 degrees from 28 to 29 GHz with a maximum aperture efficiency of 22%. This superior performance provides tantalizing evidence that the multiple switches per bit approach should not be discarded a priori due to its apparent higher complexity.

Abstract
Wireless communications in the THz frequency band is an envisioned revolutionary technology for sixth Generation (6G) networks. However, such frequencies impose certain coverage and device design challenges that need to be efficiently overcome. To this end, the development of cost- and energy-efficient approaches for scaling these networks to realistic scenarios constitute a necessity. Among the recent research trends contributing to these objectives belongs the technology of Reconfigurable Intelligent Surfaces (RISs). In fact, several high-level descriptions of THz systems based on RISs have been populating the literature. Nevertheless, hardware implementations of those systems are still very scarce, and not at the scale intended for most envisioned THz scenarios. In this paper, we overview some of the most significant hardware design and signal processing challenges with THz RISs, and present a preliminary analysis of their impact on the overall link budget and system performance, conducted in the framework of the ongoing TERRAMETA project.

Abstract
Reconfigurable Intelligent Surfaces (RISs) are in significant focus within 6G research. However, RISs face a power consumption challenge in the reconfigurable elements which may restrict its future scale-up to large areas. We address this issue by proposing a unit cell based on a non-volatile memristor-based switching mechanism. A 1-bit memristor-based reconfigurable RIS unit cell was designed in the Ka-band, and validated using CST and HFSS simulation platforms. The required control circuit to enable the digital control of the memristor has also been proposed. The proposed unit cell achieves losses of less than 1 dB over a frequency band of 25 - 28.3 GHz and a phase difference of 180 degrees +/- 20 degrees at a central frequency of 26.7 GHz, with an operational bandwidth of approximately 1 GHz. Furthermore, an exemplary 16x16 RIS was designed and simulated based on the proposed unit cell to demonstrate its capability to achieve beam steering.

Abstract
The proliferation of explicit material online, particularly pornography, has emerged as a paramount concern in our society. While state-of-the-art pornography detection models already show some promising results, their decision-making processes are often opaque, raising ethical issues. This study focuses on uncovering the decision-making process of such models, specifically fine-tuned convolutional neural networks and transformer architectures. We compare various explainability techniques to illuminate the limitations, potential improvements, and ethical implications of using these algorithms. Results show that models trained on diverse and dynamic datasets tend to have more robustness and generalisability when compared to models trained on static datasets. Additionally, transformer models demonstrate superior performance and generalisation compared to convolutional ones. Furthermore, we implemented a privacy-preserving framework during explanation retrieval, which contributes to developing secure and ethically sound biometric applications. © 2024 IEEE.

Abstract
Federated Learning (FL) is emerging in the medical field to address the need for diverse datasets while complying with data protection regulations. This decentralised learning paradigm allows hospitals (clients) to train machine learning models locally, ensuring that patient data remains within the confines of its originating institution. Nonetheless, FL by itself is not enough to guarantee privacy, as the central aggregation process may still be susceptible to identity-exposing attacks, potentially compromising data protection compliance. To strengthen privacy, differential privacy (DP) is often introduced. In this work, we conduct a comprehensive comparative analysis to evaluate the impact of DP in both traditional Centralised Learning (CL) frameworks and FL for polyp segmentation, a common medical image analysis task. Experiments are performed in PolypGen, a multi-centre publicly available dataset designed for polyp segmentation. The results show a clear drop in performance with the introduction of DP, exposing the trade-off between privacy and performance and highlighting the need to develop novel privacy-preserving techniques.

Abstract
We explore deep generative models to generate case-based explanations in a medical federated learning setting. Explaining AI model decisions through case-based interpretability is paramount to increasing trust and allowing widespread adoption of AI in clinical practice. However, medical AI training paradigms are shifting towards federated learning settings in order to comply with data protection regulations. In a federated scenario, past data is inaccessible to the current user. Thus, we use a deep generative model to generate synthetic examples that protect privacy and explain decisions. Our proof-of-concept focuses on pleural effusion diagnosis and uses publicly available Chest X-ray data. © 2024 IEEE.