Search Results (137)

Technical debt (TD) has emerged as a significant concern in the development of AI/ML applications, where rapid experimentation, evolving objectives, and complex data pipelines often introduce hidden quality and maintainability issues. Within this broader context, AI/ML competition platforms face heightened risks due to time-constrained environments and evolving requirements. Despite its relevance, TD in such competitive settings remains underexplored and lacks systematic investigation. This study addresses two research questions: (RQ1) What are the most significant types of technical debt recorded in AI-based systems? and (RQ2) How can we measure the technical debt of an AI-based competition platform? We present a scoping review of 100 peer-reviewed publications related to AI/ML competitions, aiming to map the landscape of TD manifestations and management practices. Through thematic analysis, the study identifies 18 distinct types of technical debt, each accompanied by a definition, rationale, and example grounded in competition scenarios. Based on this typology, a stakeholder-oriented assessment framework is proposed, including a detailed questionnaire and a methodology for the quantitative evaluation of TD across multiple categories. A novel contribution is the introduction of Accessibility Debt, which addresses the challenges associated with the ease and speed of immediate use of the AI/ML competition platforms. The review also incorporates bibliometric insights, revealing the fragmented and uneven treatment of TD across the literature. The findings offer a unified conceptual foundation for future work and provide practical tools for both organizers and participants to systematically detect, interpret, and address technical debt in competitive AI settings, ultimately promoting more sustainable and trustworthy AI research environments. Full article

The emergence of 6G wireless networks presents new challenges, for which cell-free massive MIMO combined with machine learning (ML) offers a promising solution. A key requirement for practical deployment is the generalizability of ML models—their ability to maintain robust performance across varying propagation conditions, user distributions, and network topologies. However, achieving generalizability typically demands large, diverse training datasets and high model complexity, which can hinder practical feasibility. This study analyzes the robustness of a low-complexity deep neural network (DNN) trained for power control under a single network configuration. The model’s robustness is assessed by testing it across a wide range of unseen scenarios, including changes in the number of access points, user equipment, and propagation environments. The DNN is trained to emulate three power control schemes: max-min spectral efficiency (SE) fairness, sum SE maximization, and fractional power control. To rigorously evaluate robustness, we compare the cumulative distribution functions of performance metrics quantitatively using the Kolmogorov–Smirnov test. Results show strong robustness, particularly for the sum SE scheme, with D statistics below 0.05 and p-values above 0.001. This work provides a reproducible framework and dataset to support further research into practical ML-based power control in cell-free massive MIMO systems. Full article

To address the issues of poor differentiation capability for high-order signals and low average recognition rates in existing communication modulation recognition techniques, this paper first performs denoising using an entropy-based dynamic Singular Value Decomposition (SVD) method and proposes a three-channel convolutional gated recurrent units (GRU) model combined with an improved SE attention mechanism for automatic modulation recognition.The model denoises in-phase/quadrature (I/Q) signals using the SVD method to enhance signal quality. By combining one-dimensional (1D) convolutional and two-dimensional (2D) convolutional, it employs a three-channel approach to extract spatial features and capture local correlations. GRU is utilized to capture temporal sequence features so as to enhance the perception of dynamic changes. Additionally, an improved SE block is introduced to optimize feature representation, adaptively adjust channel weights, and improve classification performance. Experiments on the RadioML2016.10a dataset show that the model has a maximum classification recognition rate of 92.54%. Compared with traditional CNN, ResNet, CLDNN, GRU2, DAE, and LSTM2, the average recognition accuracy is improved by 5.41% to 8.93%. At the same time, the model significantly enhances the differentiation capability between 16QAM and 64QAM, reducing the average confusion probability by 27.70% to 39.40%. Full article

Recent advances in Artificial Intelligence (AI)-driven wireless communication are driving the adoption of Sixth Generation (6G) technologies in crucial environments such as hospitals. Visible Light Communication (VLC) leverages existing lighting infrastructure to deliver high data rates while mitigating electromagnetic interference (EMI); however, patient movement induces fluctuating signal strength and dynamic channel conditions. In this paper, we present a novel integration of site-specific ray tracing and machine learning (ML) for VLC-enabled Medical Body Sensor Networks (MBSNs) channel modeling in distinct hospital settings. First, we introduce a Q-learning-based adaptive modulation scheme that meets target symbol error rates (SERs) in real time without prior environmental information. Second, we develop a Long Short-Term Memory (LSTM)-based estimator for path loss and Root Mean Square (RMS) delay spread under dynamic hospital conditions. To our knowledge, this is the first study combining ray-traced channel impulse response modeling (CIR) with ML techniques in hospital scenarios. The simulation results demonstrate that the Q-learning method consistently achieves SERs with a spectral efficiency (SE) lower than optimal near the threshold. Furthermore, LSTM estimation shows that D1 has the highest Root Mean Square Error (RMSE) for path loss (1.6797 dB) and RMS delay spread (1.0567 ns) in the Intensive Care Unit (ICU) ward, whereas D3 exhibits the highest RMSE for path loss (1.0652 dB) and RMS delay spread (0.7657 ns) in the Family-Type Patient Rooms (FTPRs) scenario, demonstrating high estimation accuracy under realistic conditions. Full article

Background/Objectives: Hospital readmissions are a key quality metric impacting both patient outcomes and healthcare costs. Traditional logistic regression models, including the LACE index (Length of stay, Admission type, Comorbidity index, and recent Emergency department visits), are commonly used for readmission risk stratification, though their accuracy may be limited by non-linear interactions with other clinical variables. This study compared the predictive performance of non-linear machine learning (ML) models with stepwise logistic regression (LR) and the LACE index for predicting 30-day general medicine readmissions. Methods: We retrospectively analysed adult general medical admissions at a tertiary hospital in Australia from 1 July 2022 to 30 June 2023. Thirty-two variables were extracted from electronic medical records, including demographics, comorbidities, prior healthcare use, socioeconomic status (SES), laboratory data, and frailty (measured by the Hospital Frailty Risk Score). Predictive models included stepwise LR and four ML algorithms: Least Absolute Shrinkage and Selection Operator (LASSO), random forest, Extreme Gradient Boosting (XGBoost), and artificial neural networks (ANNs). Performance was assessed using the area under the curve (AUC), with comparisons made using DeLong’s test. Results: Of 5371 admissions, 1024 (19.1%) resulted in 30-day readmissions. Readmitted patients were older and frailer and had more comorbidities and lower SES. Logistic regression (LR) identified the key predictors of outcomes, including heart failure, alcoholism, nursing home residency, and prior admissions, achieving an AUC of 0.62. LR’s performance was comparable to that of the LACE index (AUC = 0.61) and machine learning models: LASSO (AUC = 0.63), random forest (AUC = 0.60), and artificial neural networks (ANNs) (AUC = 0.60) (p > 0.05). However, LR significantly outperformed XGBoost (AUC = 0.55) (p < 0.05). Conclusions: About one in five general medicine patients are readmitted within 30 days. Traditional LR performed as well as or better than ML models for readmission risk prediction. Full article

In this paper, we present a practical partitioned intelligent-reflecting-surface-aided transmit spatial modulation (PIRS-TSM) scheme, where spatial modulation is implemented at the transmitter and partitioning is conducted on the IRS to enhance the spectral efficiency (SE) and reliability for multiple-input single-output (MISO) systems. The theoretical analysis of average bit error rate (ABER) based on maximum likelihood (ML) detection and the computational complexity analysis are provided. Experimental simulations demonstrate that the PIRS-TSM scheme obtains a significant ABER enhancement under the same SE compared to the existing partitioned IRS-aided transmit space shift keying or generalized space shift keying schemes by additionally carrying modulated symbols. Moreover, the system performances with different configurations of antenna numbers and symbol modulation orders under the same SE are investigated as a practical application reference. Full article

(1): Background: This study aimed to evaluate the concentrations of four proteins for allergic patch testing (APT) in dogs, assessing sensitivity (SE), specificity (SP), negative predictive value (NPV), positive predictive value (PPV), reactions to adhesives/containers, and the safety of APT with food proteins in dogs. (2) Methods: For evaluation, 43 dogs were screened and divided into two groups: Group 1 consisted of 20 healthy dogs, and Group 2 included 23 dogs with canine atopic dermatitis (AD). Group 1 underwent allergic patch testing (APT) with beef, pork, chicken, and soy proteins at four different concentrations (100 mg, 250 mg, 500 mg, 1000 mg/0.2 mL). Of the 23 dogs included in Group 2, four did not undergo the elimination diet and were excluded, leaving 17 dogs in the study. They underwent an elimination diet (ED) and were evaluated using the pruritus visual analog scale (pVAS) and lesion scores (CADESI-4) before and after the ED (days 0 and 45). After the ED, Group 2 was subjected to APT (using the same proteins and concentrations as Group 1) and an oral provocation test (OPT) with the proteins used in the APT. The results of the OPT were used to assess the accuracy of the APT. (3) Results: In Group 1, one dog reacted to the APT. In Group 2, after 45 days of ED, of the 17 dogs included, 13 showed a reduction in pVAS and CADESI-4 scores (p < 0.05) and nine an improvement considered good to excellent. Of these, two showed irritant contact reactions to the APT chambers and were excluded, leaving 11 dogs that were reactive to APT, and the OPT increased pruritus (p < 0.05). Accuracy: Beef and chicken proteins at concentrations of 500 and 1000 mg/0.2 mL, and soy protein at 1000 mg/0.2 mL, achieved 100% SE, SP, PPV, and NPV. Pork protein at 1000 mg/0.2 mL achieved 100% SE, 83% SP, 83% PPV, and 100% NPV. (4) Conclusions: APT with beef and chicken proteins at 500 mg and 1000 mg/0.2 mL and soy protein at 1000 mg/0.2 mL, based on the results of this study, can be recommended for diagnosing adverse food reactions in dogs with AD. Full article

Maritime communication networks are critical for supporting the increasing demands of oceanic and coastal activities, including shipping, fishing, and offshore operations. However, traditional systems face significant challenges in providing reliable, high-throughput connectivity due to dynamic sea environments, mobility, and non-line-of-sight (NLoS) conditions. Reconfigurable intelligent surfaces (RISs) have been proposed as a promising solution to overcome these limitations by enabling programmable control of electromagnetic wave propagation in next-generation mobile communication networks, such as beyond fifth generation and sixth generation ones (B5G/6G). This paper presents a deep learning-based (DL) scheme for beam selection in RIS-aided maritime next-generation networks. The proposed approach leverages deep learning to optimize beam selection dynamically, enhancing signal quality, coverage, and network efficiency in complex maritime environments. By integrating RIS configurations with data-driven insights, the proposed framework adapts to changing channel conditions and potential vessel mobility while minimizing latency and computational overhead. Simulation results demonstrate significant improvements in both machine learning (ML) metrics, such as beam selection accuracy, and overall communication reliability compared to traditional methods. More specifically, the proposed scheme reaches around 99% Top-$K$ Accuracy levels while jointly improving energy efficiency (ee) and spectral efficiency (SE) by approx. 2 times compared to state-of-the-art approaches. This study provides a robust foundation for employing DL in RIS-aided maritime networks, contributing to the advancement of intelligent, high-performance wireless communication systems for advanced maritime applications, such as autonomous mooring, the autonomous approach, and just-in-time arrival (JIT). Full article

Marine litter (ML), encompassing human-made objects in marine ecosystems, poses significant threats to the coasts of some Adriatic islands, despite their remoteness and sparse populations. These islands, reliant on tourism, are particularly vulnerable to ML pollution. This study hypothesized that the natural features of the islands influence ML distribution. It employes an integrated geographic approach combining the results of field survey (via sea kayaking) with various indicators which include: (1) coastal orientation and number density of bays, (2) vegetation exposure and biomass share, (3) island area and number density of bays, (4) bay openness and ML quantity, and (5) bay openness and plastic prevalence in ML. Focusing on islands of Lošinj, Pašman, Vis, and the Kornati and Elaphiti archipelago, the study analyzed data collected over six years (2018–2023). Results highlighted that NW-SE and W-E coastal orientations are particularly susceptible to ML accumulation, especially in the southern Adriatic. Linear Fitting Regression analyses revealed a stronger correlation between number density of polluted bays and the surface area of smaller islands (<10 km²) compared to larger islands (>10 km²). The following findings underscore the need for international collaboration and stringent policies to mitigate ML pollution, ensuring the protection of Adriatic marine ecosystems and the sustainability of local communities. Full article

The progress of artificial intelligence (AI), particularly its core algorithms—machine learning (ML) and deep learning (DL)—has been significant in the medical field, impacting both scientific research and clinical practice. These algorithms are now capable of analyzing ultrasound images, processing them, and providing outcomes, such as determining the benignity or malignancy of thyroid nodules. This integration into ultrasound machines is referred to as computer-aided diagnosis (CAD). The use of such software extends beyond ultrasound to include cytopathological and molecular assessments, enhancing the estimation of malignancy risk. AI’s considerable potential in cancer diagnosis and prevention is evident. This article provides an overview of AI models based on ML and DL algorithms used in thyroid diagnostics. Recent studies demonstrate their effectiveness and diagnostic role in ultrasound, pathology, and molecular fields. Notable advancements include content-based image retrieval (CBIR), enhanced saliency CBIR (SE-CBIR), Restore-Generative Adversarial Networks (GANs), and Vision Transformers (ViTs). These new algorithms show remarkable results, indicating their potential as diagnostic and prognostic tools for thyroid pathology. The future trend points to these AI systems becoming the preferred choice for thyroid diagnostics. Full article

Advanced strategies for production forecasting, operational optimization, and decision-making enhancement have been employed through reservoir management and machine learning (ML) techniques. A hybrid model is established to predict future gas output in a gas reservoir through historical production data, including reservoir pressure, cumulative gas production, and cumulative water production for 67 months. The procedure starts with data preprocessing and applies seasonal exponential smoothing (SES) to capture seasonality and trends in production data, while an Artificial Neural Network (ANN) captures complicated spatiotemporal connections. The history replication in the models is quantified for accuracy through metric keys such as mean absolute error (MAE), root mean square error (RMSE), and R-squared. The future forecast is compared with an outcome of a previous physical model that integrates wells and reservoir properties to simulate gas production using regressions and forecasts based on empirical and theoretical relationships. Regression analysis ensures alignment between historical data and model predictions, forming a baseline for hybrid model performance evaluation. The results reveal the complementary attributes of these methodologies, providing insights into integrating data-driven and physics-based approaches for optimal reservoir management. The hybrid model captured the production rate conservatively with an extra margin of three years in favor of the physical model. Full article

Magnetic ionic liquid (MIL) based on alkyl phosphonium cation was used as a curing agent for developing epoxy nanocomposites (ENCs) modified with a graphene nanoplatelet (GNP)/carbon nanotube (CNT) hybrid filler. The materials were prepared by a solvent-free procedure involving ball-milling technology. ENCs containing as low as 3 phr of filler (GNP/CNT = 2.5:0.5 phr) exhibited electrical conductivity with approximately six orders of magnitude greater than the system loaded with GNP = 2.5 phr. Moreover, the use of MIL (10 phr) resulted in ENCs with higher conductivity compared with the same system cured using conventional aliphatic amine. The filler dispersion within the epoxy matrix was confirmed by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). The electromagnetic interference shielding effectiveness (EMI SE), evaluated in the X- and Ku-band frequency range, revealed a great contribution of the absorption mechanism for the ENC containing the hybrid filler and cured with MIL. Moreover, the best microwave-absorbing response was achieved with the ENC containing GNP/CNT = 2.5/0.5 phr, and cured with ML, which a minimum RL of −23.61 dB and an effective absorption bandwidth of 5.18 GHz were observed for thickness of 1.5 mm. In summary, this system is a promising material for both civilian and military applications due to its simple and scalable nanocomposite preparation method, the lightweight nature of the composites resulting from the low filler content, the commercial availability and cost-effectiveness of GNP, and its high electromagnetic wave attenuation across a broad frequency range. Full article

The emergence of 5G technology promises remarkable advancements in wireless communication, particularly in the realm of mmWave (millimeter-wave) massive multiple input multiple output (m-MIMO) systems. However, the realization of its full potential is hindered by the challenge of pilot overhead, which compromises system efficiency. The efficient usage of pilot signals is crucial for precise channel estimation and interference reduction to maintain data integrity. Nevertheless, this requirement brings up the challenge of pilot overhead, which utilizes precious spectrum space, thus reducing spectral efficiency (SE). To address this obstacle, researchers have progressively turned to artificial intelligence (AI) and machine learning (ML) methods to design hybrid beam-forming systems that enhance SE while reducing changes to the bit error rate (BER). This study addresses the challenge of pilot overhead in hybrid beamforming for 5G mmWave m-MIMO systems by leveraging advanced artificial intelligence (AI) techniques. We propose a framework integrating k-clustering, linear regression, random forest regression, and neural networks with singular value decomposition (NN-SVD) to optimize pilot placement and hybrid beamforming strategies. The results demonstrate an 82% reduction in pilot overhead, a 250% improvement in spectral efficiency, and a tenfold enhancement in bit error rate at low SNR conditions, surpassing state-of-the-art methods. These findings validate the efficacy of the proposed system in advancing next-generation wireless networks. Full article

Background: Pembrolizumab is a monoclonal antibody (mAb) approved for treating Non-Small Cell Lung Cancer (NSCLC), melanoma and lymphomas. Commercialized in single-size (100 mg/4 mL) vials, the pembrolizumab solution contains no preservative. As such, the manufacturer recommends using pembrolizumab vials only once, and thus, to rapidly dispose of any unused portion. Thus, appreciable amounts of this costly product are wasted. Objective: To evaluate the physical, chemical and microbiological stability of pembrolizumab vial leftovers stored at room temperature or at 4 °C, 7 and 14 days after first vial puncturing. Methods: Following pH assessments, submicronic aggregation and turbidity of pembrolizumab were measured by dynamic light scattering (DLS) and spectrophotometry, respectively. In addition, SE-HPLC (size-exclusion high-performance liquid chromatography), IEX-HPLC (ion exchange HPLC) and peptide mapping HPLC served to respectively evaluate aggregation and fragmentation, distribution of charge and primary structure of pembrolizumab. Incubation at 37 °C for 48 h of pembrolizumab vial leftovers on blood agar plates was used to determine their microbiological stability. Results: Physical, chemical and microbiological stability of pembrolizumab leftovers was demonstrated for at least two full weeks. Conclusions: These results argue forcefully in favor of allowing prolongation of pembrolizumab vial leftovers usage well beyond a single day. Full article

Step emulsification (SE) is renowned for its robustness in generating monodisperse emulsion droplets at arrayed nozzles. However, few studies have explored poly(dimethylsiloxane) (PDMS)-based SE devices for producing monodisperse oil-in-water (O/W) droplets and polymeric microspheres with diameters below 20 µm—materials with broad applicability. In this study, we present a PDMS-based microfluidic SE device designed to achieve this goal. Two devices with 264 nozzles each were fabricated, featuring straight and triangular nozzle configurations, both with a height of 4 µm and a minimum width of 10 µm. The devices were rendered hydrophilic via oxygen plasma treatment. A photocurable acrylate monomer served as the dispersed phase, while an aqueous polyvinyl alcohol solution acted as the continuous phase. The straight nozzles produced polydisperse droplets with diameters exceeding 30 µm and coefficient-of-variation (CV) values above 10%. In contrast, the triangular nozzles, with an opening width of 38 µm, consistently generated monodisperse droplets with diameters below 20 µm, CVs below 4%, and a maximum throughput of 0.5 mL h⁻¹. Off-chip photopolymerization of these droplets yielded monodisperse acrylic microspheres. The low-cost, disposable, and scalable PDMS-based SE device offers significant potential for applications spanning from laboratory-scale research to industrial-scale particle manufacturing. Full article