Abstract
Covid-19 is a severe illness caused by the Sars-CoV-2 virus, initially identified in China in late 2019 and swiftly spreading globally. Since the virus primarily impacts the lungs, analyzing chest X-rays stands as a reliable and widely accessible means of diagnosing the infection. In computer vision, deep learning models such as CNNs have been the main adopted approach for detection of Covid-19 in chest X-ray images. However, we believe that handcrafted features can also provide relevant results, as shown previously in similar image classification challenges. In this study, we propose a method for identifying Covid-19 in chest X-ray images by extracting and classifying local and global percolation-based features. This technique was tested on three datasets: one comprising 2,002 segmented samples categorized into two groups (Covid-19 and Healthy); another with 1,125 non-segmented samples categorized into three groups (Covid-19, Healthy, and Pneumonia); and a third one composed of 4,809 non-segmented images representing three classes (Covid-19, Healthy, and Pneumonia). Then, 48 percolation features were extracted and give as input into six distinct classifiers. Subsequently, the AUC and accuracy metrics were assessed. We used the 10-fold cross-validation approach and evaluated lesion sub-types via binary and multiclass classification using the Hermite polynomial classifier, a novel approach in this domain. The Hermite polynomial classifier exhibited the most promising outcomes compared to five other machine learning algorithms, wherein the best obtained values for accuracy and AUC were 98.72% and 0.9917, respectively. We also evaluated the influence of noise in the features and in the classification accuracy. These results, based in the integration of percolation features with the Hermite polynomial, hold the potential for enhancing lesion detection and supporting clinicians in their diagnostic endeavors.

Abstract
Spectral imaging is a broad term that refers to the use of a spectroscopy technique to analyze sample surfaces, collecting and representing spatially referenced signals. Depending on the technique utilized, it allows the user to reveal features and properties of objects that are invisible to the human eye, such as chemical or molecular composition. However, the interpretability and interaction with the results are often limited to screen visualization of two-dimensional representations. To surpass such limitations, augmented reality emerges as a promising technology, assisted by recent developments in the integration of spectral imaging datasets onto three-dimensional models. Building on this context, this work explores the integration of spectral imaging with augmented reality, aiming to create an immersive toolset to increase the interpretability and interactivity of the results of spectral imaging analysis. The procedure follows a two-step approach, starting from the integration of spectral maps onto a three-dimensional models, and proceeding with the development of an interactive interface to allow immersive visualization and interaction with the results. The approach and tool developed present the opportunity for a user-centric extension of reality, enabling more intuitive and comprehensive analyses with the potential to drive advancements in various research domains.

Abstract

Abstract
The electrification of the transport sector is a critical element in the transition to a low-emissions economy, driven by the widespread adoption of electric vehicles (EV) and the integration of renewable energy sources (RES). However, managing the increasing demand for EV charging infrastructure while meeting carbon emission reduction targets is a significant challenge for charging station operators. This work introduces a novel carbon-aware dynamic pricing framework for EV charging, formulated as a chance-constrained optimization problem to consider forecast uncertainties in RES generation, load, and grid carbon intensity. The model generates day-ahead dynamic tariffs for EV drivers with a certain elastic behavior while optimizing costs and complying with a carbon emissions budget. Different types of budgets for Scope 2 emissions (indirect emissions of purchased electricity consumed by a company) are conceptualized and demonstrate the advantages of a stochastic approach over deterministic models in managing emissions under forecast uncertainty, improving the reduction rate of emissions per feasible day of optimization by 24 %. Additionally, a surrogate machine learning model is proposed to approximate the outcomes of stochastic optimization, enabling the application of state-of-the-art explainability techniques to enhance understanding and communication of dynamic pricing decisions under forecast uncertainty. It was found that lower tariffs are explained, for instance, by periods of higher renewable energy availability and lower market prices and that the most important feature was the hour of the day.

Abstract
This study presents the development and testing of satellite antennas for the SONDA probe, an innovative deepsea monitoring system designed to be deployed by high-altitude balloons. The probe descends to the deep ocean, resurfaces, and transmits data while functioning as a drifter. The project faced unique design constraints, including the need for low-cost materials and lightweight construction for balloon deployment. These constraints ruled out traditional hermetic housings, necessitating alternative solutions for antenna protection. The work focused on custom ceramic patch antennas and their performance under various protective coatings, which affected the antennas' resonance and gain. Thinner layers effectively protected the antennas from high-pressure conditions and water ingress, maintaining functionality. Experiments on antenna height revealed optimal positioning above the water surface to minimize wave-induced signal interference. Hyperbaric chamber tests validated the mechanical integrity and functionality of the antennas under pressures equivalent to depths of 1500 m Antenna characterization techniques were employed in an anechoic chamber to validate antenna performance with the coating and to assess their correct operation after the hyperbaric tests. Field deployments demonstrated the antennas' capability to transmit data after diving. Challenges included communication delays, corrupted data, and mechanical vulnerabilities in materials. The findings emphasize the importance of rigorous mechanical design, material selection, and system optimization to ensure reliability in marine environments. This work advances the development of low-cost, lightweight, and modular probes for autonomous ocean monitoring, with potential applications in long-term drifter studies, real-time marine monitoring and oceanographic research.

Abstract
Hyperparameter optimization (HPO) is critical for enhancing the predictive performance of machine learning models in credit risk assessment for peer-to-peer (P2P) lending. This study evaluates four HPO methods, Grid Search, Random Search, Hyperopt, and Optuna, across four models, Logistic Regression, Random Forest, XGBoost, and LightGBM, using three real-world datasets (Lending Club, Australia, Taiwan). We assess predictive accuracy (AUC, Sensitivity, Specificity, G-Mean), computational efficiency, robustness, and interpretability. LightGBM achieves the highest AUC (e.g., 70.77% on Lending Club, 93.25% on Australia, 77.85% on Taiwan), with XGBoost performing comparably. Bayesian methods (Hyperopt, Optuna) match or approach Grid Search’s accuracy while reducing runtime by up to 75.7-fold (e.g., 3.19 vs. 241.47 min for LightGBM on Lending Club). A sensitivity analysis confirms robust hyperparameter configurations, with AUC variations typically below 0.4% under ±10% perturbations. A feature importance analysis, using gain and SHAP metrics, identifies debt-to-income ratio and employment title as key default predictors, with stable rankings (Spearman correlation > 0.95, p<0.01) across tuning methods, enhancing model interpretability. Operational impact depends on data quality, scalable infrastructure, fairness audits for features like employment title, and stakeholder collaboration to ensure compliance with regulations like the EU AI Act and U.S. Equal Credit Opportunity Act. These findings advocate Bayesian HPO and ensemble models in P2P lending, offering scalable, transparent, and fair solutions for default prediction, with future research suggested to explore advanced resampling, cost-sensitive metrics, and feature interactions.