Delivering Training and Education to 200.000 Employees by means of Advanced Learning Architectures

The aim of the paper is to present and compare available standards describing requirements for quality management in training companies.Two standards were taken into consideration: international standard ISO 29990 and the Malopolska Standards for Education and Training Services (MSUES) which has been developed by the Regional Labour Office in Cracow (Poland).Studies of this type of standards are very scarce in the literature.Most of the publications about training concern the quality of training from the view of a customer, effectiveness of training methods or design of training activities, but not the quality management in training company.The study proves that this kind of standard can have a positive impact on the market of training services.The study also identifies the key similarities and differences among the analysed standards.

The aim of this work is to show that organizational change is a response to certain problems generated by a market specialization strategy. Our study is based on two cases of SMEs that first engaged in a business relationship with a single client for more than five years. This led to a high degree of skills specialization and a certain organizational rigidity which was subsequently detrimental to the continuity of the company's activity, particularly when the relationship with the only single client was unexpectedly concluded, for the client had become increasingly demanding in terms of cost and time. Analysis of the data shows that both SMEs have made a series of organizational changes (restructuring of the company, recruitment of new skills, training activities, use of partnerships with other organizations, etc.) to overcome this perilous situation. As a consequence, these companies were able to reorient their business policy in order to reduce their dependence and then ensure their business sustainability and growth. Keywords: Organizational change, dependency situation, customer-supplier relationship, strategic choices. DOI: 10.7176/EJBM/12-27-04 Publication date: September 30 th 2020

Hogarth recently identified that what was mainly missing in the evaluation of management training was a model of what was intended to be achieved in the first place. If we have no clear idea what effects are sought, the most sophisticated measurement techniques are to no avail. Foy also argued recently that more boundary crossing was needed in the 1980s between Business Schools and company training departments. If we put their two well substantiated propositions together, we have a need for a model and a programme of activities in management training which straddles both the Business School's resources and the company's learning environment. Visually, this is traditionally accomplished by a Venn diagram, and in Figure 1 this is attempted.

Boosting efficiency in state estimation of power systems by leveraging attention mechanism

In recent years, there are many great successes in using deep architectures for unsupervised feature learning from data, especially for images and speech. In this paper, we introduce recent advanced deep learning models to classify two emotional categories (positive and negative) from EEG data. We train a deep belief network (DBN) with differential entropy features extracted from multichannel EEG as input. A hidden markov model (HMM) is integrated to accurately capture a more reliable emotional stage switching. We also compare the performance of the deep models to KNN, SVM and Graph regularized Extreme Learning Machine (GELM). The average accuracies of DBN-HMM, DBN, GELM, SVM, and KNN in our experiments are 87.62%, 86.91%, 85.67%, 84.08%, and 69.66%, respectively. Our experimental results show that the DBN and DBN-HMM models improve the accuracy of EEG-based emotion classification in comparison with the state-of-the-art methods.

This research is based on a systematic review of machine learning (ML) approaches for the cognitive load (CL) assessment of applications for unmanned aerial system (UAS) operator training. The review synthesises evidence on how ML techniques have been applied to assess CL using diverse data sources, including physiological signals (e.g., EEG, HRV), behavioural measures (e.g., eye-tracking), and performance indicators. It highlights the effectiveness of models such as Support Vector Machines (SVMs), Random Forests (RFs), and advanced deep learning (DL) architectures such as Long Short-Term Memory (LSTM), as well as how the use of different methods affects the performance of ML models, with studies reporting accuracies of up to 98%. The findings also indicate that, compared with traditional UAS training approaches, ML approaches can enhance training by providing adaptive assessment, with methodological factors such as model selection, data preprocessing, and validation being central to ML assessment performance. These findings highlight the value of accurate CL assessment as a foundation for adaptive training systems, supporting enhanced UAS operator performance and operational safety. By consolidating the methodological insights and identifying research gaps, this review provides valuable background information for advancing ML-based CL assessment and its integration into adaptive UAS operator training systems to enhance UAS operator training.

In precision manufacturing, the selection of cutting tools and uncontrolled tool wear is a major challenge which directly affects manufactured product quality, surface integrity and process efficiency. This work uses advanced deep learning (DL) architectures: InceptionNetv3, ResNeXt, and InceptionResNetv2 that have been tuned through metaheuristic optimization strategies to monitor and choose milling cutters. The Tunable Q Wavelet Transform (TQWT) is used in the proposed approach to preprocess and convert vibration signals to scalograms, which act as an input feature for deep learning models. To further improve prediction accuracy, each model’s hyperparameters are adjusted using Ant Lion Optimization (ALO) and the Clonal Selection Algorithm (CSA). Experimental investigations on milling dataset demonstrate that, the CSA-optimized InceptionResNetv2 exhibited the average prediction accuracy (98.25%), according to five-fold cross-validation for cutter c4. The ALO and CSA tuned models demonstrate improved generalization through five-fold cross validations ensuring robustness under varying operating and cutter conditions. The proposed approach offers a computationally efficient and scalable solution for real time cutter selection and tool wear monitoring in industrial environments. By eliminating conventional feature extraction and enabling adaptive learning from vibration based scalograms, the proposed framework can be integrated in Industry 4.0 manufacturing systems for predictive maintenance and process optimization.

Large earthquakes (EQs) occur at surprising loci and timing, and their descriptions remain a long-standing enigma. Finding answers by traditional approaches or recently emerging machine learning (ML)-driven approaches is formidably difficult due to data scarcity, interwoven multiple physics, and absent first principles. This paper develops a novel artificial intelligence (AI) framework that can transform raw observational EQ data into ML-friendly new features via basic physics and mathematics and that can self-evolve in a direction to better reproduce short-term large EQs. An advanced reinforcement learning (RL) architecture is placed at the highest level to achieve self-evolution. It incorporates transparent ML models to reproduce magnitude and spatial location of large EQs (\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$M_w \\ge $$\\end{document} 6.5) weeks before of the failure. Verifications with 40-year EQs in the western U.S. and comparisons against a popular EQ forecasting method are promising. This work will add a new dimension of AI technologies to large EQ research. The developed AI framework will help establish a new database of all EQs in terms of ML-friendly new features and continue to self-evolve in a direction of better reproducing large EQs.

In recent years, Deep Learning (DL) models, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Long Short-Term Memory (LSTM), have been widely used for Human Activities Recognition (HAR). They have achieved considerable performance improvements over classical Machine Learning (ML) approaches due to their excellent feature representation capabilities. Simple human activities are performed in sequential order with no overlaps or concurrent actions. The paucity of labeled training activity samples, high computational cost, and system resource requirements of deep learning architectures compared to lightweight models are some of the research challenges confronting the HAR community. We framed the lightweight DL-based CNNs, RNNs, and LSTM model for HAR to tackle these research challenges. The RNNs and LSTM have single layers with softmax activation functions, whereas the one-dimensional CNNs have two convolutional and single max-pooling layers. To evaluate the performance of our models, we used the publicly available benchmark WISDM experimental dataset, which includes the reading of six activities (walking, jogging, upstairs, downstairs, sitting, and standing) performed by 36 participants using a single accelerometer sensor. The experimental result illustrates the efficacy of our models in activity recognition, demonstrating that they attain higher accuracy while being computationally efficient.

Abdominal Aortic Aneurysm (AAA) remains a significant public health challenge, with an 82.1% increase in related fatalities from 1990 to 2019. In the United States alone, AAA complications resulted in an estimated 13,640 deaths between 2018 and 2021. In clinical practice, computed tomography angiography (CTA) is the primary imaging modality for monitoring and pre-surgical planning of AAA patients. CTA provides high-resolution vascular imaging, enabling detailed assessments of aneurysm morphology and informing critical clinical decisions. However, manual segmentation of CTA images is labor-intensive and time consuming, underscoring the need for automated segmentation algorithms, particularly when feature extraction from clinical images can inform treatment decisions. We propose a framework to automatically segment the outer wall of the abdominal aorta from CTA images and estimate AAA wall stress. Our approach employs a patch-based dilated modified U-Net model to accurately delineate the outer wall boundary of AAAs and Nonlinear Elastic Membrane Analysis (NEMA) to estimate their wall stress. We further integrate Non-Uniform Rational B-Splines (NURBS) to refine the segmentation. During prediction, our deep learning architecture requires 17±0.02 milliseconds per frame to generate the final segmented output. The latter is used to provide critical insight into the biomechanical state of stress of an AAA. This modeling strategy merges advanced deep learning architecture, the precision of NURBS, and the advantages of NEMA to deliver a robust and efficient method for computational analysis of AAAs.

The accurate prediction of sports consumption behavior is vital for market precision and resource optimization. While multi-modal data availability has grown, existing methods face significant challenges in fusing heterogeneous data sources and establishing causal relationships for marketing interventions. To address these gaps, this study proposes HUMF-Net, a novel hybrid deep learning architecture for sports consumption potential prediction. First, by learning cross-modal dependencies of heterogeneous data, the model dynamically aligns and fuses multiple data sources, including user profiles, consumption history, online search behavior, and social media interactions. Then, the fused features are processed by a hierarchical temporal feature extraction network (HTF-Net) to extract unified feature representations. This encoder synergistically combines CNNs to capture short-term local patterns and transient peaks, and a dynamic feature enhancement network (DFE-Net) to model long-term behavioral sequences and periodicity, effectively learning multi-scale temporal dependencies in consumption behavior. Finally, a dedicated uplift module maps these deep representations to consumption propensity scores, producing the final prediction output. Experimental results demonstrate that on a real-world sports consumption dataset, our model achieves a 53.3% conversion rate improvement over advanced deep learning methods, with a prediction accuracy of 96.2% and an MAPE is 2.1%.

Medical image captioning has emerged as a critical intersection of computer vision and natural language processing, offering significant promise for enhancing clinical radiology workflows. This paper proposes a method for generating diagnostic descriptions from chest X-ray images using advanced deep learning architectures. Three methodologies are evaluated: a conventional encoder-decoder baseline, a model integrating CheXNet with custom flow modules, and an attention-guided framework that emphasizes relevant anatomical regions. The parallel Global Flow and Context Flow pathways are designed to capture both comprehensive anatomical context and localized pathological details essential for accurate diagnosis. Evaluation results indicate that the attention-based model outperforms the alternatives, demonstrating improved capability in identifying clinically relevant features while adhering to radiological language standards. Despite these advancements, challenges remain, including dataset imbalance toward normal findings and limitations in detecting rare pathologies. These findings offer important direction for the development of AI systems that support radiological interpretation while maintaining diagnostic accuracy and clinical appropriateness.

Environmental pollution is an important common health concern across the world with patterns of exposure being quite spatial and time varying. Traditional measures of pollution and epidemiological research tends to use sparse monitoring systems and single studies, which restrict their ability to resolve dynamic effects of exposure and health at large scale. To overcome these shortcomings, this research is one of the deep learning based multimodal frameworks that incorporate remote sensing information and biomedical health records in the overall spatio-temporal analysis of environmental pollution and health risk prediction. Pollution indicators of aerosol optical depth and trace gas concentrations collected by satellites are used together with anonymized biomedical information (incidence of diseases and hospital admission history) with a deep learning architecture, which is an attention competition. The suggested model will use convolutional neural networks to learn the spatial pollution features, and temporal modeling networks to learn the dynamics of evolving exposure factors, the fusion mechanism is based on the attention, the cross-modal relationships between the environment and health data will be learned. Real-world experimental analyses on experiment datasets show the proposed multimodal approach is always better than the traditional single-modal, and statistical analysis, with a higher rate of pollution estimation and better rate of health risk prediction. The maps of spatial heat analyses also demonstrate that the model is successful in determining regions of high risks and the most exposing time periods. The findings prove that the combination of heterogeneous environmental and biomedical data with the help of advanced deep learning procedures can be used to make more sensible and valuable assessments of environmental health. The article identifies the possibility of multimodal artificial intelligence systems to assist in active environmental health monitoring, early signs of danger, and evidence based approach to policymaking in the management of sustainable human public health.

Pneumonia is a serious illness that poses a major health risk to children worldwide, and early detection is crucial for effective treatment. This study explores a range of machine learning and deep learning methods for accurately identifying pneumonia from pediatric chest X-ray images. To improve image quality, preprocessing techniques such as adaptive histogram equalization and normalization were applied. Several models were evaluated, including traditional Convolutional Neural Networks (CNNs) and advanced transfer learning architectures like VGG16, ResNet50, and EfficientNetB0. Among all models, EfficientNetB0 achieved the highest accuracy of 96.4%, demonstrating strong potential for assisting clinical diagnosis. The study also highlights that carefully designed preprocessing and optimized model selection can significantly enhance detection performance while keeping computational requirements manageable. These findings suggest that lightweight, automated approaches can provide practical support to healthcare professionals, enabling faster and more reliable pneumonia detection in children.

The integration of Artificial Intelligence and Machine Learning technologies into Computer-Aided Design systems represents a transformative approach to addressing longstanding challenges in engineering design processes. Traditional metadata-based search methods have proven inadequate for efficiently locating existing components, resulting in significant economic losses across manufacturing sectors. Shape-based search emerges as a compelling alternative, leveraging advanced deep learning architectures to enable intuitive geometric similarity matching. This capability fundamentally alters how engineers interact with design repositories, allowing for component retrieval based on visual similarity rather than textual descriptions. The implementation of shape-based search yields substantial benefits, including dramatic reductions in search time, increased component reuse rates, and enhanced design standardization. While integration challenges exist, organizations successfully deploying these technologies report compelling return on investment through reduced development cycles and lower certification costs. As computational technologies continue to advance, the application of geometric deep learning to CAD search promises to further revolutionize engineering knowledge management by enabling cross-domain component discovery and function-based retrieval capabilities.