Abstract
Radon (Rn-222) is a unique atmospheric tracer, since it is an inert gaseous radionuclide with a predominantly terrestrial source and a short half-life (3.8232 (8) d), enabling quantification of the relative degree of recent (< 21 d) terrestrial influences on marine air masses. High quality measurements of atmospheric radon activity concentration in remote oceanic locations enable the most accurate identification of baseline conditions. Observations of GHGs under baseline conditions, representative of hemispheric background values, are essential to characterise long-term changes in hemispheric-mean GHG concentrations, differentiate between natural and anthropogenic GHG sources, and improve understanding of the global carbon budget.The EU-funded project NuClim (Nuclear observations to improve Climate research and GHG emission estimates) will establish world-leading high-quality atmospheric measurements of radon activity concentration and of selected GHG concentrations (CO2, and CH4) at a remote oceanic location, the Eastern North Atlantic (ENA) facility, managed by the Atmospheric Radiation Measurement (ARM) programme (Office of Science from the U.S. Department of Energy), located on Graciosa Island (Azores archipelago), near the middle of the north Atlantic Ocean. These observations will provide an accurate, time-varying atmospheric baseline reference for European greenhouse gas (GHG) levels, enabling a clearer distinction between anthropogenic emissions and slowly changing background levels. NuClim will also enhance measurement of atmospheric radon activity concentration at the Mace Head Station, allowing the identification of latitudinal gradients in baseline atmospheric composition, and supporting the evaluation of the performance of GHG mitigation measures for countries in the northern hemisphere.The high-quality nuclear and GHG observations from NuClim, and the resulting classification of terrestrial influences on marine air masses, will assist diverse climate and environmental studies, including the study of pollution events, characterisation of marine boundary layer clouds and aerosols, and exploration of the impact of natural planktonic communities on GHG emissions. This poster presents an overview of NuClim, outlines the project objectives and methodologies, and summarises the relevant data products that will be made available to the climate community.Project NuClim received funding from the EURATOM research and training program 2023-2025 under Grant Agreement No 101166515.

Abstract

Abstract
This study presents an explainable content-based image retrieval system for predicting post-surgical aesthetic outcomes in breast cancer patients, comparing state-of-theart vision transformers, convolutional neural networks, and B-cos architectures. Results show that vision transformers, particularly GC ViT and DaViT, outperform convolutional neural networks and B-cos architectures, achieving an adjusted discounted cumulative gain of up to 80.18%. This superior performance is attributed to their ability to model long-range dependencies while effectively capturing local information. Bcos networks underperform (64.28-70.19% adjusted discounted cumulative gain), likely due to oversimplified feature alignment unsuitable for clinical tasks. Explainability analysis using Integrated Gradients reveals that models primarily focus on breast regions but occasionally attend to irrelevant features (e.g., arm positioning, leading to retrieval errors and highlighting a semantic gap between learned visual similarities and clinical relevance. Future work aims to integrate anatomical segmentation and ensemble learning methods to enhance clinical alignment and address attention inaccuracies. Clinical Relevance-The content-based image retrieval system developed in this study aids clinicians by supporting surgical outcome prediction in breast cancer patients and streamlining the traditionally time-intensive task of manually identifying similar reference images for patient consultation. © 2025 IEEE.

Abstract
We introduce a multimodal Signal Quality Indicator (SQI) for assessing fidelity of synchronous electrocardiogram (ECG) and phonocardiogram (PCG) signals recorded in ambulatory, non-standardized settings. The method uses a bidirectional fiducial-matching algorithm to test the temporal alignment of QRS complexes and T waves (ECG) with S1 and S2 sounds (PCG) respectively. Validation employed 564 synchronous ECG–PCG pairs collected with the FDA-cleared Rijuven Cardiosleeve at the aortic, pulmonary, tricuspid, and mitral valves sites. Expert annotations served as ground truth. In a three-class task, the SQI reached an area under the ROC curve greater than 79%, showing strong discriminative power. This physiology-based metric supports batch-online monitoring and reliable quality control of opportunistic cardiac recordings.

Abstract
The Digital Services Act (DSA) establishes a regulatory framework for online platforms and search engines in the European Union, focusing on mitigating systemic risks such as illegal content dissemination, fundamental rights violations, and impacts on electoral processes, public health, and gender-based violence. Very Large Online Platforms (VLOPs) and Very Large Search Engines (VLOSEs), defined as those with over 45 million active recipients, must provide data access for research to enable investigations into these risks and the development of solutions. This tutorial is tailored for the UMAP community, addressing the implications of the DSA for user modelling research. It will cover the DSA's key provisions and definitions, outline the procedural steps for accessing VLOP and VLOSE data, and discuss the technical aspects of data access requests. Participants will also explore the challenges and opportunities involved in working with this data. By the end of the tutorial, attendees will have a thorough understanding of the DSA's data access provisions, the technical and procedural requirements for accessing VLOP and VLOSE data, and the regulation's implications for user modelling research. They will be equipped to navigate the complexities of the DSA and contribute to the development of responsible and transparent online platforms.Further information and resources about the tutorial are available on the website: https://erasmopurif.com/tutorial-dsa-umap25/.

Abstract
The main goal of this paper was to develop, implement, and test a practical framework for large-scale last-mile delivery problems that employ a combination of optimisation and machine learning while focussing on different routing methods. Delivery companies in big cities choose delivery orders based on the tacit knowledge of experienced drivers, since solving a large optimisation model with several variables is not a practical solution to meet their daily needs. This framework includes three phases of districting, sequencing, and routing, and in total 30 different variants were tested in different capacities. Using the power of machine learning, a model is trained and tuned to predict driving road distances, allowing the implementation of the whole framework and improving performance from analysing 2983 stops in several hours to 58,192 stops in less than 15 minutes. The results demonstrated that Inter 1 - Centroids is the best inter-district connection method, and one of the best variants in this framework is variant 26 which managed to decrease up to 34,77% total distances with 79 fewer drivers in a full month analysis compared to the original routes of the delivery company. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.