Abstract

Abstract
In many real-world prediction tasks, the class labels contain information about the relative order between the labels that are not captured by commonly used loss functions such as multicategory cross-entropy. In ordinal regression, many works have incorporated ordinality into models and loss functions by promoting unimodality of the probability output. However, current approaches are based on heuristics, particularly nonparametric ones, which are still insufficiently explored in the literature. We analyze the set of unimodal distributions in the probability simplex, establishing fundamental properties and giving new perspectives to understand the ordinal regression problem. Two contributions are then proposed to incorporate the preference for unimodal distributions into the predictive model: 1) UnimodalNet, a new architecture that by construction ensures the output is a unimodal distribution, and 2) Wasserstein regularization, a new loss term that relies on the notion of projection in a set to promote unimodality. Experiments show that the new architecture achieves top performance, while the proposed new loss term is very competitive while maintaining high unimodality.

Abstract
The evolution of markets provides a change in the way organisations act. To improve their competitive performance and stay on the market, organisations often adopt a strategy to establish agreements with other organisations, known as strategic alliances. Several tools, algorithms, and computational systems call upon other sciences as a source of inspiration. In this work we explore flocking behaviour, a paradigm of biology, to analyse the collective intelligence behaviour that emerges from a group of individuals or firms. Inspired by the Cucker and Smale algorithm (C-S), we propose a new version of the flocking algorithm, AllFlock, applied to strategic alliances, considering a learning mechanism. For this new approach, metrics were obtained for the parameters of the C-S algorithm: position, velocity, and influence. The latter uses cooperative games, adapted mechanisms, and methods currently explored in reinforcement learning. We have used Netlogo as the modelling environment. Five parameter configurations were analysed. For each of those configurations, the average number of iterations, the permanence rate of organisations in the alliance, and the average growth of the organisations were computed. The behaviour of the organisations reveals a tendency for convergence, confirming the existence of flocking behaviour. © 2025 The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG.

Abstract
This paper addresses two challenges that are intertwined and are key in informing signal processing methods restoring natural (voiced) speech from whispered speech. The first challenge involves characterizing and modeling the evolution of the harmonic phase/magnitude structure of a sequence of individual pitch periods in a voiced region of natural speech comprising sustained or co-articulated vowels. A novel algorithm segmenting individual pitch pulses is proposed, which is then used to obtain illustrative results highlighting important differences between sustained and co-articulated vowels, and suggesting practical synthetic voicing approaches. The second challenge involves model-based synthetic voicing restoration in real-time and on-the-fly. Three implementation alternatives are described that differ in their signal reconstruction approaches: frequency-domain, combined frequency- and time-domain, and physiologically inspired filtering of glottal excitation pulses individually generated. The three alternatives are compared objectively using illustrative examples, and subjectively using the results of listening tests involving synthetic voicing of sustained and co-articulated vowels in word context. © 2025 Elsevier B.V., All rights reserved.

Abstract
Background Cardiovascular disease (CVD) is a major cause of mortality worldwide, necessitating more refined strategies for risk assessment. Recently, lipoprotein(a) [Lp(a)] has gained attention for its distinctive role in atherosclerosis, yet its prevalence and impact for cardiovascular risk assessment are not well-documented in the Portuguese population. This study aimed to characterize Lp(a) levels in a real-world Portuguese cohort, investigating its prevalence and association with CVD risk. Methods Retrospective and cross-sectional study of adults who underwent serum Lp(a) analysis in a Portuguese hospital between August 2018 and June 2022. Demographic and anthropometric data, laboratory values, relevant comorbidities and lipid-lowering medication were collected. Results Of 1134 participants, 28.7% had elevated Lp(a) levels (> 125 nmol/L). A higher prevalence was observed in those with atherosclerotic cardiovascular disease (ASCVD) (45.9%) or a family history of premature CVD (41.9%). Additionally, a significant association was found between elevated Lp(a) levels and traditional CVD risk factors, including hypertension, dyslipidemia, and diabetes mellitus. Among those classified as having low-to-moderate CVD risk by (Systematic COronary Risk Evaluation 2) SCORE2, 55.7% exhibited high Lp(a) levels (> 75 nmol/L), suggesting a potential higher risk of CVD disease. Conclusions The prevalence of elevated Lp(a) in Portugal, notably among those with ASCVD or premature CVD history, is concerning. This study underscores the potential of Lp(a) assessment for a more comprehensive approach to cardiovascular risk assessment. This could improve the stratification of CVD risk and identify individuals who could benefit from early intensive management of their risk factors, ultimately reducing the burden of CVD and cardiovascular-related mortality.

Abstract
This research addresses the challenge of generating synthetic data that resembles real-world data while preserving privacy. With privacy laws protecting sensitive information such as healthcare data, accessing sufficient training data becomes difficult, resulting in an increased difficulty in training Machine Learning models and in overall worst models. Recently, there has been an increased interest in the usage of Generative Adversarial Networks (GAN) to generate synthetic data since they enable researchers to generate more data to train their models. GANs, however, may not be suitable for privacy-sensitive data since they have no concern for the privacy of the generated data. We propose modifying the known Conditional Tabular GAN (CTGAN) model by incorporating a privacy-aware loss function, thus resulting in the Private CTGAN (PCTGAN) method. Several experiments were carried out using 10 public domain classification datasets and comparing PCTGAN with CTGAN and the state-of-the-art privacy-preserving model, the Differential Privacy CTGAN (DP-CTGAN). The results demonstrated that PCTGAN enables users to fine-tune the privacy fidelity trade-off by leveraging parameters, as well as that if desired, a higher level of privacy.