30793 research outputs found
Sort by
Fed-Hetero: A Self-Evaluating Federated Learning Framework for Data Heterogeneity
Get PDFData Availability Statement:
Data are contained within the article.Federated learning (FL) enables deep learning models to be trained locally on devices without the need for data sharing, ensuring data privacy. However, when clients have uneven or imbalanced data distributions, it leads to data heterogeneity. Data heterogeneity can appear in different ways, often due to variations in label, data distributions, feature variations, and structural inconsistencies in the images. This can significantly impact FL performance, as the global model often struggles to achieve optimal convergence. To enhance training efficiency and model performance, a common strategy in FL is to exclude clients with limited data. However, excluding such clients can raise fairness concerns, particularly for smaller populations. To understand the influence of data heterogeneity, a self-evaluating federated learning framework for heterogeneity, Fed-Hetero, was designed to assess the type of heterogeneity associated with the clients and provide recommendations to clients to enhance the global model’s accuracy. Fed-Hetero thus enables the clients with limited data to participate in FL processes by adopting appropriate strategies that enhance model accuracy. The results show that Fed-Hetero identifies the client with heterogeneity and provides personalized recommendations.This research received no external funding
XAI-DSCSA: explainable-AI-based deep semi-supervised convolutional sparse autoencoder for facial expression recognition
No full textData availability:
No datasets were generated or analysed during the current study.Acknowledgements:
The authors sincerely thank the ISO Certified (ISO/IEC 20000-1:2018) Centre for Machine Learning and Intelligence (CMLI), funded by the Department of Science and Technology (DST-CURIE), India, for providing the facility to carry out this research study.Facial expression recognition (FER) continues to be a vibrant research field, driven by the increasing need for its practical applications in areas such as e-learning, healthcare, candidate interview analysis, and more. Most deep learning approaches in supervised FER systems heavily rely on large, labeled datasets. Implementing FER in Convolutional Neural Networks (CNNs) often requires many layers, leading to extended training times and difficulties in finding optimal parameters. This can result in challenges in creating distinct facial expression patterns for classification, leading to poor real-time emotion classification In this paper, we propose a novel approach known as the Deep Semi-supervised Convolutional Sparse Autoencoder to address the aforementioned issues and enhance FER performance and prediction accuracy. This approach comprises two parts: (i) Initially, a deep convolutional sparse autoencoder is trained with unlabeled samples of facial expressions. Here, sparsity is introduced in the convolutional block to enforce penalties, focusing on more relevant features for feature representation in the latent space. (ii) A trained encoder with a feature map is connected to a fully connected layer with softmax for final fine-tuning with learned weights and labeled facial expression samples in a semi-supervised approach for emotion classification. This approach was experimented with two benchmark datasets, namely CK + and JAFFE, and achieved significant results of 98.98% and 93.10% accuracy, respectively. The results were analyzed using established state-of-the-art techniques. Additionally, eXplainable AI (XAI) methods like Grad-CAM and image-LIME were employed to interpret the performance and prediction outcomes of the DSCSA model.This research study received no external funding
Pulsed reverse electrochemical synthesis of Ag-TiO2 composites from deep eutectic solvents: Photocatalytic and antibacterial behaviour
Get PDFData availability:
No data was used for the research described in the article.Supplementary materials are available online at: https://www.sciencedirect.com/science/article/pii/S2666523925000571#sec0015 .This study presents an environmentally friendly approach for synthesis Ag-TiO2 composite using pulsed reverse current (PRC) electrodeposition from green electrolytes, specifically deep eutectic solvents (DESs). The combination of PRC and DESs offers better control over nanoparticle synthesis while eliminating the need for toxic or expensive precursors, representing a significant advancement in sustainable nanomaterial synthesis. Different electrochemical parameters were adjusted, and their influence on the structure and morphology of the composite was investigated using X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM) and transmission electron microscopy (TEM). TEM analysis revealed that silver nanoparticles (Ag NPs) are attached to TiO2 nanopowder, with the coexistence of TiO2 and Ag further confirmed by XRD and XPS. The recorded UV–Vis diffuse reflectance spectra (DRS) displayed a broad peak in the range of 400 – 650 nm, associated with the localized surface plasmon resonance (LSPR) of Ag NPs on the semiconductor’s surface. The photocatalytic activity of TiO2 nanopowder and Ag-TiO2 composite was evaluated based on the degradation of methyl orange (MO) dye under UV and visible light illumination. Our findings clearly demonstrated that the incorporation of Ag improved the photocatalytic efficiency. The mechanism of MO dye degradation was explored by using various scavengers, revealing that superoxide radicals (•O−
2 ) play a dominant role. Furthermore, the incorporation of Ag NPs significantly enhanced the antimicrobial activity of the oxide against both Gram-positive (B. subtilis) and Gram-negative (E.coli) strains.The present work was supported by the Romanian National Grant GNAC ARUT 2023 project (No. 7/06.10.2023), NANO_NP_DES, entitled "Electrochemical synthesis of hybrid nanoparticles Ag/TiO2 and Ag/Fe3O4 with biomedical applications". Also, the work was supported by the IMT Core Program µNanoEl, within the PNCDI 2022-2026, carried out with the support of Romanian Ministry of Research, Innovation and Digitization, project No. 23070201.
N.D. acknowledges the support by the contract PN 23 21 01 06 sponsored by the Romanian Ministry of Research, Innovation, and Digitization.
Additionally, National and European funds financially supported the work through FCT under Research Grant UIDB/00081/2020 (https://doi.org/10.54499/UIDB/00081/2020)- CIQUP, LA/P/0056/2020 (https://doi.org/10.54499/LA/P/0056/2020) – IMS
Towards a sustainable and circular blue bioeconomy: A scoping review
Get PDFData availability:
Data will be made available on request.This study explores biotechnological innovations based on marine resources in the burgeoning blue bioeconomy, aiming to unlock their potential for valorization for sustainable and circular business models. Using a scoping review to identify key trends in scientific knowledge, we specify four clusters of blue biotechnological innovations: (1) bioenergy, (2) feedstock and fertilizers, (3) biomass for food, and (4) industrial applications. These four clusters offer promising energy, food, and materials alternatives while aligning with sustainability and circular economy principles. This study thus fills a gap in extant management research by integrating ocean biodiversity into considerations of sustainable entrepreneurship and investigating valorization options in the blue bioeconomy for the development of novel business models. The findings pave the way for designing feedstock and regenerative business models that harness the blue bioeconomy's economic, environmental, and social benefits
Digital twin role for sustainable and resilient renewable power plants: A systematic literature review
Get PDFData availability:
No data was used for the research described in the article.Supplementary data are available online at: https://www.sciencedirect.com/science/article/pii/S2213138825000281?via%3Dihub#appSB .Transitioning to sustainable and resilient energy generation presents challenges in optimizing resource and storage utilization, reducing operational costs, and addressing environmental impacts within renewable energy power plants. The shift away from fossil fuels in the energy sector requires innovative solutions to enhance sustainability and resilience. This study aims to explore the role of Digital Twin (DT) technology – a digital replica of a physical object or process with bidirectional communication – in promoting sustainability within power plants, an area that remains underexplored. Using a Sytematic Literature Review (SLR) of 61 peer-reviewed papers, this research examines six key categories of DT application: predictive analysis, performance optimization, risk assessment, model evaluation, process traceability, and human–machine interaction. The findings indicate that DT holds significant potential to improve power plant sustainability by enabling cost reductions, optimizing energy usage, and minimizing environmental impact through waste reduction and carbon emission management. This study underscores DT’s importance in supporting the energy sector’s transition towards sustainable practices and enhancing the resilience of renewable energy systems.Waqar Ali Khan received PhD scholarships from Higher Education Commission and Swinburne University of Technology
Robust deep dictionary learning via self-expression neighbor atom enhancement
No full textData availability:
Data will be made available on request.Dictionary learning has been widely recognized as an effective method for processing high-dimensional and nonlinear data. Although graph structures have been widely adopted to enhance optimization in irregular data handling, existing approaches inadequately utilize local graph information, leading to suboptimal dictionary update. Consequently, the learned dictionary struggles to capture local details effectively, which can negatively impact the performance of classification tasks. To tackle this problem, this study proposes a method based on graph structure, dubbed Deep Neighbor Atom-enhanced Dictionary Learning (DNADL), designed to enhance the sensitivity of dictionaries to local structure. Our DNADL segments data into internally consistent neighborhood manifold module via clustering method and enhances dictionary atoms based on their self-expression relationships, where the self-expression relationships within generated manifolds reflect the intrinsic correlation of local data structures. Enhanced dictionary atoms exploit the similarities among neighboring data points, ensuring that the constructed dictionary effectively captures local structure features. Furthermore, this study introduces an adaptive graph learning module that dynamically updates graph representations, thereby facilitating the simultaneous optimization of dictionary learning and graph topology within an integrated framework. Comprehensive experimental results demonstrate that DNADL achieves state-of-the-art classification performance across multiple benchmark datasets, with accuracies of 99.23 % on Fashion-MNIST, 94.12 % on EMNIST, and 95.3 % on Oxford Flowers 102. By effectively capturing local structural features, DNADL surpasses existing methods by a significant margin of 0.98 % to 3.70 % in accuracy.This work was supported by the National Natural Science Foundation of China under Grants 62472201, 62172193 and 62376108
Personalized Federated Learning for GAI-Assisted Semantic Communications
Get PDFSemantic Communication (SC), which focuses on transmitting meaning rather than raw data, has emerged as the next-generation communication paradigm. However, the performance of SC is heavily influenced by network design and training methodologies. To address these challenges and enhance SC performance at the edge, we first introduce a Generative Artificial Intelligence (GAI)-assisted SC (GSC) model, which improves SC capabilities by optimizing the network architecture. Then, to achieve the efficient learning of GSC models deployed on each user, a Personalized Semantic Federated Learning (PSFL) framework is proposed. Specifically, in the local training phase, a Personalized Local Distillation (PLD) approach is employed, where each user selects a tailored GSC model as a mentor based on local resources. This mentor subsequently distills knowledge to a unified student model, ensuring compliance with the model isomorphism requirements of FL. In the global aggregation phase, an Adaptive Global Pruning (AGP) scheme is applied, dynamically pruning or expanding the aggregated global model based on real-time channel conditions. This mechanism effectively balances accuracy and communication energy efficiency. Finally, numerical results validate the feasibility and efficacy of the proposed PSFL framework, demonstrating its potential to enhance SC performance in edge environments significantly.Tpaper was partly funded by Jiangsu Major Project on Basic Research (Grant No.: BK20243059), Gusu Innovation Project for People (Grant No.: ZXL2024360), Natural Science Foundation of China (Grant No. 62132004), the Major Program Project of Xiangjiang Laboratory (Grant No. XJ2023001 and XJ2022001), and Qiyuan Lab Innovation Fund (Grant No. 2022-JCJQ-LA-001-088)
Subsurface seawalls for sustainable coastal aquifers and inundation prevention under climate change and tidal waves impact
Get PDFData availability statement:
Upon request.Code availability:
Upon request.This study investigates saltwater intrusion (SWI) dynamics in the Biscayne aquifer, which includes two scenarios (1) simulation scenarios were developed using SEAWAT to assess the aquifer’s response to sea level rise (SLR) and tidal fluctuation-induced pressure waves at the base case (before using seawall) and (2) using seawall intervention to mitigate SWI and SLR flooding. SLR projections were modeled using boundary conditions informed by National Oceanic and Atmospheric Administration (NOAA) and The Intergovernmental Panel on Climate Change (IPCC) data, revealing progressive SWI advancement up to 8900 m intrusion under +2.62 m SLR compared to the baseline where the SWI extent was 7950 m. Tidal impacts, modeled using observed Broward County tidal ranges and an amplified sinusoidal (SIN) wave, showed intrusion length up to 8750 m for the tidal fluctuations value of +2.31 m. Seawall height requirements to manage overtopping were calculated using empirical wave run-up models, indicating a minimum design height of 7.5 m to resist a worst-case SLR of +2.62 m and maximum tide of +2.31 m. The findings underscore the importance of adaptive coastal management, integrating engineering solutions with hydrogeologic and ecological considerations.This study did not receive any funding
Implementation of Machine – Learning Models to study Habitat Suitability for Deep-Diving Cetaceans in the Ikarian Basin Paper
No full textThe authors would like to thank the Archipelagos Institute of Marine Conservation team for their support with data collection.The Ikarian Basin, located in the eastern Mediterranean Sea, is home to three deep-diving cetacean species: sperm whales (Physeter macrocephalus), goose-beaked whales (Ziphius cavirostris), and Risso’s dolphins (Grampus griseus). These species, classified as endangered or vulnerable in the Mediterranean by the IUCN, face multiple threats, including overexploitation of fish stocks, habitat degradation, and the effects of climate change. Understanding their biology and ecology - migrations, habitat use, behavioural patterns, and residency dynamics - is essential for the development of effective conservation strategies. Modelling is a key tool for predicting species movements and habitat suitability. This study analyses the habitat suitability of 41 observations of Grampus griseus, 87 of Physeter macrocephalus, and 95 of Ziphius cavirostris, using depth, slope, and distance from shore as predictor variables. Using a Random Forest model, species were accurately classified, achieving an out-of-bag (OOB) error rate of 10.76%. Results indicate all environmental variables significantly influence species occurrence. These findings underscore the importance of the Ikarian Basin as a suitable habitat for deepdiving cetaceans, providing a foundation for proposing the area as a potential Important Marine Mammal Area (IMMA). Further research incorporating logged behavioural data and acoustic detections is needed to refine conservation strategies and gain deeper insights into habitat utilization
Optimizing rotational velocity for melting performance of heat storage tank containing metal foam in building heating system
No full textData availability:
No data was used for the research described in the article.Phase change thermal energy storage (TES) represents a crucial technology for enhancing the efficiency of solar energy utilization, and optimizing the heat transfer performance of TES units has attracted significant interest. In this investigation, a computational model of a horizontal tube-and-shell TES unit was used to analyze heat transfer performance under both active and passive enhancement methods involving the integration of metal foam and the application of rotational conditions, and a rotating TES experimental platform was constructed to verify the numerical model. Meanwhile, the charging performance of composite phase change material (CPCM) under various rotational speeds was investigated. Optimal rotational conditions were selected by comparing parameters such as complete melting time, heat storage rate, heat storage capacity, temperature response rate, and liquid phase, temperature, and velocity distributions. The results indicate that the inclusion of metal foam improves the heat storage efficiency of the phase change material (PCM), resulting in a reduction of the total melting time for CPCM device by a factor of 36 compared with pure PCM device under stationary condition. Additionally, the melting rate of the rotational setup is improved compared to the stationary setup, particularly for CPCM units. Subsequently, a detailed analysis of TES units under various rotational speeds revealed that as the rotational speed increases, the complete melting time decreases. Specifically, at a rotational speed of 0.6 rpm, the melting time decreases by 12 %. Moreover, the rotational mechanism also enhances the temperature uniformity within the TES unit. Notably, further increasing the rotational speed beyond 0.6 rpm does not alter the complete melting time.This work was supported by the Key Scientific and Technological Innovation Team of Shaanxi Province (2023-CX-TD-29)
Data sources:
Access Repository Dashboard
Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇