Abstract
This study investigates the use of SHAP (SHapley Additive exPlanations) values as an explainable artificial intelligence (xAI) technique applied on a facial attribute classification task. We analyse the consistency of SHAP value distributions across diverse classifier architectures that share the same feature extractor, revealing that key features driving attribute classification remain stable regardless of classifier architecture. Our findings highlight the challenges in interpreting SHAP values at the individual sample level, as their reliability depends on the model's ability to learn distinct class-specific features; models exploiting inter-class correlations yield less representative SHAP explanations. Furthermore, pixel-level SHAP analysis reveals that superior classification accuracy does not necessarily equate to meaningful semantic understanding; notably, despite FaceNet exhibiting lower performance than CLIP, it demonstrated a more nuanced grasp of the underlying class attributes. Finally, we address the computational scalability of SHAP, demonstrating that KernelExplainer becomes infeasible for high-dimensional pixel data, whereas DeepExplainer and GradientExplainer offer more practical alternatives with trade-offs. Our results suggest that SHAP is most effective for small to medium feature sets, providing interpretable and computationally manageable explanations.

Abstract
The investigation of business performance via the lens of sustainability has become an increasingly attractive topic among scholars. This study contributes to the field by proposing a two-stage methodology. First, to assess companies' efficiency in terms of sustainability, their scores of the environmental, social and governance (ESG) pillars are combined into the single weighted sustainability performance indicator using the 'Benefit of the Doubt' model, which is maximized for each company by comparing it against the best performers in terms of ESG scores based on Data Envelopment Analysis. Then, in the second stage, the significant determinants are identified after efficiency estimates are regressed on company performance indicators using Tobit panel regression. To demonstrate this approach, we selected 559 companies from the manufacturing sector, as this industry continues to face challenges to reduce environmental impact, improve resource efficiency, and promote social responsibility. The main findings include the examination of the best performers and underperforming companies in terms of sustainability, along with key financial indicators identified in the study.

Abstract
BackgroundDialysis recovery time (DRT) refers to the period during which fatigue and weakness subside following hemodialysis treatment, allowing patients to resume their daily routines. This study aimed to identify the factors influencing DRT in hemodialysis patients in Turkey and Portugal, where the prevalence of chronic kidney disease is notably high.MethodsA cross-sectional observational study was conducted in a private dialysis center in Turkey and three dialysis centers in Portugal. The study included hemodialysis patients aged 18 years or older who had been undergoing four-hour hemodialysis sessions three times a week for at least six months. Participants had no communication barriers and voluntarily agreed to take part in the study. Data were collected using a semi-structured questionnaire to gather descriptive characteristics and the Hospital Anxiety and Depression Scale. Logistic regression analysis was employed to identify independent variables influencing DRT.ResultsA total of 294 patients participated in the study, including 187 from Turkey and 107 from Portugal. In Turkey, increased interdialytic weight gain (P = 0.043) was associated with prolonged recovery time, while the use of high-flux dialyzers (P = 0.026) was linked to shorter recovery times. In Portugal, older age (P = 0.020) was found to extend recovery time.ConclusionRecovery time after dialysis is influenced by varying factors across different countries. Further research with larger sample sizes is needed to deepen understanding of these factors and their implications.Clinical trial numberNCT04667741.

Abstract
The growing integration of intermittent renewable energy sources poses new challenges to power system stability [...]

Abstract
A semi-distributed optical fiber bending extensometer system based on OTDR is proposed, consisting of N-loops designed to enable different maximum extension measurements and sensitivities. This system offers a low-cost solution for monitoring landslides and similar civil structures. Tests conducted at 1625 nm demonstrate that different series of sensors can be independently measured with elongation errors typically within +/- 0.25 cm across a range from 0 to 9 cm.

Abstract
Software businesses are continuously increasing their presence in the cloud. While cloud computing is not a new research topic, designing software for the cloud is still challenging, requiring engineers to invest in research to become proficient at working with it. Design patterns can be used to facilitate cloud adoption, as they provide valuable design knowledge and implementation guidelines for recurrent engineering problems. This work introduces a pattern language for designing software for the cloud. We believe developers can significantly reduce their R&D time by adopting these patterns to bootstrap their cloud architecture. The language comprises 10 patterns, organized into four categories: Automated Infrastructure Management, Orchestration and Supervision, Monitoring, and Discovery and Communication.