Abstract
Background/Objectives: Type 2 diabetes mellitus (T2DM) is a chronic condition marked by hyperglycemia, which can affect metabolic, vascular, and hematological parameters. A low-carbohydrate, high-fat (LCHF) diet has been shown to improve glycemic control and blood pressure regulation. Exercise in hypoxia (EH) enhances insulin sensitivity, erythropoiesis, and angiogenesis. The combination of LCHF and EH may offer a promising strategy for managing T2DM and hypertension (HTN), although evidence remains limited. This study aimed to assess the effects of an eight-week normobaric EH intervention at 3000 m simulated altitude combined with an LCHF diet on hematological and lipid profiles, inflammation, and blood pressure in older patients with T2DM and HTN. Methods: Forty-two diabetic patients with HTN were randomly assigned to three groups: (1) control group (control diet + exercise in normoxia), (2) EH group (control diet + EH), and (3) intervention group (EH+LCHF) Baseline and eight-week measurements included systolic, diastolic, and mean blood pressure (SBP, DBP, MAP), hematological and lipid profiles, and inflammation biomarkers. Results: Blood pressure decreased after the intervention (p < 0.001), with no significant differences between groups (SBP: p = 0.151; DBP: p = 0.124; MAP: p = 0.18). No differences were observed in lipid profile or C-reactive protein levels (p > 0.05). Mean corpuscular hemoglobin (MCH) increased in the EH group (p = 0.027), while it decreased in the EH+LCHF group (p = 0.046). Conclusions: Adding hypoxia or restricting carbohydrates did not provide additional benefits on blood pressure in T2DM patients with HTN. Further elucidation of the mechanisms underlying hematological adaptations is imperative.

Abstract

Abstract
The large gap between the generalization level of state-of-the-art machine learning and human learning systems calls for the development of artificial intelligence (AI) models that are truly inspired by human cognition. In tasks related to image analysis, searching for pixel-level regularities has reached a power of information extraction still far from what humans capture with image-based observations. This leads to poor generalization when even small shifts occur at the level of the observations. We explore a perspective on this problem that is directed to learning the generative process with causality-related foundations, using models capable of combining symbolic manipulation, probabilistic reasoning, and pattern recognition abilities. We briefly review and explore connections of research from machine learning, cognitive science, and related fields of human behavior to support our perspective for the direction to more robust and human-like artificial learning systems.

Abstract
The need for real-time and scalable oceanographic monitoring has become crucial for coastal management, marine traffic control and environmental sustainability. This study investigates the integration of sensor technology into marine cables to enable real-time monitoring, focusing on tidal cycles and wave characteristics. A 2000 m cable demonstrator was deployed off the coast of Portugal, featuring three active repeater nodes equipped with pressure sensors at varying depths. The goal was to estimate hourly wave periods using fast Fourier transform and calculate significant wave height via a custom peak detection algorithm. The results showed strong coherence with tidal depth variations, with wave period estimates closely aligning with forecasts. The wave height estimations exhibited a clear relationship with tidal cycles, which demonstrates the system's sensitivity to coastal hydrodynamics, a factor that numerical models designed for open waters often fail to capture. The study also highlights challenges in deep-water monitoring, such as signal attenuation and the need for high sampling rates. Overall, this research emphasises the scalability of sensor-integrated smart marine cables, offering a transformative opportunity to expand oceanographic monitoring capabilities. The findings open the door for future real-time ocean monitoring systems that can deliver valuable insights for coastal management, environmental monitoring and scientific research.

Abstract
Many current AutoML platforms include a very large space of alternatives (the configuration space). This increases the probability of including the best one for any dataset but makes the task of identifying it for a new dataset more difficult. In this paper, we explore a method that can reduce a large configuration space to a significantly smaller one and so help to reduce the search time for the potentially best algorithm configuration, with limited risk of significant loss of predictive performance. We empirically validate the method with a large set of alternatives based on five ML algorithms with different sets of hyperparameters and one preprocessing method (feature selection). Our results show that it is possible to reduce the given search space by more than one order of magnitude, from a few thousands to a few hundred items. After reduction, the search for the best algorithm configuration is about one order of magnitude faster than on the original space without significant loss in predictive performance.

Abstract