Scalable Model Research Articles

To Push Forward the Integrated Reform in An All-Round Way and Focus on Building A Strong Province in The Digital Economy

This paper, taking Dalian as a case study, delves into the strategic pathways and practical strategies for comprehensively advancing integrated reform and jointly striving to build a digital economy powerhouse province. It first analyzes the significance of the digital economy as a new engine for global economic growth, along with Dalian's foundational advantages and potential challenges in its digital transformation journey. Subsequently, from various dimensions such as policy guidance, technological innovation, industrial upgrading, and business environment optimization, the paper elaborates on how Dalian is leveraging integrated reform measures, including enhancing cross-departmental collaboration, promoting open sharing of data resources, and constructing a digital economy ecosystem, to accelerate the deep integration of digital technologies with the real economy, thereby promoting industrial structural optimization and upgrading, and enhancing regional competitiveness. The paper also emphasizes Dalian's key initiatives in nurturing and attracting talents, strengthening infrastructure support, and advancing smart city and digital government construction during its pursuit of becoming a digital economy powerhouse, and how these initiatives provide strong momentum for the high-quality development of Dalian's, and by extension, its province's, digital economy. Through case studies and practical summaries, this paper offers a series of replicable and scalable experiences and models, providing valuable insights for other regions to advance their own digital economy integration reforms.

Strategic management of CO2: A scalable model for CCS in decarbonised societies

In future decarbonised energy systems, residual carbon emissions require strategic planning and management. In environmental management, an evaluation of carbon removal considering local geographic frameworks is needed. This paper introduces a scalable and adaptable model for evaluating the economics and geography of future carbon capture and storage (CCS) configurations across geographical scales, covering capture, transport, and storage of carbon. The model is applied to the North Denmark Region, showing that future energy production carbon sources will be concentrated in Thisted and Jammerbugt, while industrial sources remain in Aalborg and Rebild municipalities. Carbon transport configurations, including truck, pipeline, and shipping are assessed, for the carbon to be stored in onshore and offshore geological storages. The regional scale findings suggest that pipelines and onshore storage provide the most economical configuration. However, a sensitivity study using a smaller geographical scope indicates potential for optimising carbon transport by evaluating both carbon volume and distance. The paper discusses how the model's flexibility and scalability enable the integration of alternate cost components, thereby supporting the calculation of the carbon repurposing potentials, including carbon capture, utilisation, and storage (CCUS) configurations.

Using machine learning to create an adaptable, scalable, and interpretable behavioral model.

Using machine learning to create an adaptable, scalable, and interpretable behavioral model.

The Ouest Data Hub: An Interregional Health Data Sharing Ecosystem for Research.

The Ouest Data Hub (ODH) a project lead by GCS HUGO which is a cooperation group of University Hospitals in the French Grand Ouest region represents a groundbreaking initiative in this territory, advancing health data sharing and reuse to support research driven by real-world health data. Central to its structure are the Clinical Data Warehouses (CDWs) and Clinical Data Centers (CDCs), essential for analytics and as the linchpin of the ODH's status as an interregional Learning Health System. Aimed at fostering innovation and research, the ODH's collaborative and multi-institutional model effectively utilizes both local and shared resources. Yet, the path is not without challenges, especially in data quality and interoperability, where ongoing harmonization and standard adherence are critical. In 2023, this facilitated access to extensive health data from over 9.3 million patient records, demonstrating the ODH's capacity for both monocentric and multicentric research across various clinical fields, in close collaboration with physicians. The integration of healthcare professionals is crucial, ensuring data's clinical relevance and guiding accurate interpretations. Future expansions of the ODH to new hospitals and data types promise to enhance its model further, already inspiring similar frameworks across France. This scalable model for health data ecosystems showcases the ODH's potential as a foundation for national and supranational data sharing efforts.

Retinoic acid induces human gastruloids with posterior embryo-like structures

Gastruloids are a powerful in vitro model of early human development. However, although elongated and composed of all three germ layers, human gastruloids do not morphologically resemble post-implantation human embryos. Here we show that an early pulse of retinoic acid (RA), together with later Matrigel, robustly induces human gastruloids with posterior embryo-like morphological structures, including a neural tube flanked by segmented somites and diverse cell types, including neural crest, neural progenitors, renal progenitors and myocytes. Through in silico staging based on single-cell RNA sequencing, we find that human RA-gastruloids progress further than other human or mouse embryo models, aligning to E9.5 mouse and CS11 cynomolgus monkey embryos. We leverage chemical and genetic perturbations of RA-gastruloids to confirm that WNT and BMP signalling regulate somite formation and neural tube length in the human context, while transcription factors TBX6 and PAX3 underpin presomitic mesoderm and neural crest, respectively. Looking forward, RA-gastruloids are a robust, scalable model for decoding early human embryogenesis.

Advancing breast ultrasound diagnostics through hybrid deep learning models

Today, doctors rely heavily on medical imaging to identify abnormalities. Proper classification of these abnormalities enables them to take informed actions, leading to early diagnosis and treatment. This paper introduces the “EfficientKNN” model, a novel hybrid deep learning approach that combines the advanced feature extraction capabilities of EfficientNetB3 with the simplicity and effectiveness of the k-Nearest Neighbors (k-NN) algorithm. Initially, EfficientNetB3, pre-trained on ImageNet, is repurposed to serve as a feature extractor. Subsequently, a GlobalAveragePooling2D layer is applied, followed by an optional Principal Component Analysis (PCA) to reduce dimensionality while preserving critical information. PCA is used selectively when deemed necessary. The extracted features are then classified using an optimized k-NN algorithm, fine-tuned through meticulous cross-validation.Our model underwent rigorous training using a curated dataset containing benign, malignant, and normal medical images. Data augmentation techniques, including rotations, shifts, flips, and zooms, were employed to help the model generalize and efficiently handle new, unseen data. To enhance the model's ability to identify the important features necessary for accurate predictions, the dataset was refined using segmentation and overlay techniques. The training utilized an ensemble of optimization algorithms—SGD, Adam, and RMSprop—with hyperparameters set at a learning rate of 0.00045, a batch size of 32, and up to 120 epochs, facilitated by early stopping to prevent overfitting.The results demonstrate that the EfficientKNN model outperforms traditional models such as VGG16, AlexNet, and VGG19 in terms of accuracy, precision, and F1-score. Additionally, the model showed better results compared to EfficientNetB3 alone. Achieving a 100 % accuracy rate on multiple tests, the EfficientKNN model has significant potential for real-world diagnostic applications. This study highlights the model's scalability, efficient use of cloud storage, and real-time prediction capabilities, all while minimizing computational demands.By integrating the strengths of EfficientNetB3's deep learning architecture with the interpretability of k-NN, EfficientKNN presents a significant advancement in medical image classification, promising improved diagnostic accuracy and clinical applicability.

Ameliorating Racial Disparities in HIV Prevention via a Nurse-Led, AI-Enhanced Program for Pre-Exposure Prophylaxis Utilization Among Black Cisgender Women: Protocol for a Mixed Methods Study.

HIV pre-exposure prophylaxis (PrEP) is a critical biomedical strategy to prevent HIV transmission among cisgender women. Despite its proven effectiveness, Black cisgender women remain significantly underrepresented throughout the PrEP care continuum, facing barriers such as limited access to care, medical mistrust, and intersectional racial or HIV stigma. Addressing these disparities is vital to improving HIV prevention outcomes within this community. On the other hand, nurse practitioners (NPs) play a pivotal role in PrEP utilization but are underrepresented due to a lack of awareness, a lack of human resources, and insufficient support. Equipped with the rapid evolution of artificial intelligence (AI) and advanced large language models, chatbots effectively facilitate health care communication and linkage to care in various domains, including HIV prevention and PrEP care. Our study harnesses NPs' holistic care capabilities and the power of AI through natural language processing algorithms, providing targeted, patient-centered facilitation for PrEP care. Our overarching goal is to create a nurse-led, stakeholder-inclusive, and AI-powered program to facilitate PrEP utilization among Black cisgender women, ultimately enhancing HIV prevention efforts in this vulnerable group in 3 phases. This project aims to mitigate health disparities and advance innovative, technology-based solutions. The study uses a mixed methods design involving semistructured interviews with key stakeholders, including 50 PrEP-eligible Black women, 10 NPs, and a community advisory board representing various socioeconomic backgrounds. The AI-powered chatbot is developed using HumanX technology and SmartBot360's Health Insurance Portability and Accountability Act-compliant framework to ensure data privacy and security. The study spans 18 months and consists of 3 phases: exploration, development, and evaluation. As of May 2024, the institutional review board protocol for phase 1 has been approved. We plan to start recruitment for Black cisgender women and NPs in September 2024, with the aim to collect information to understand their preferences regarding chatbot development. While institutional review board approval for phases 2 and 3 is still in progress, we have made significant strides in networking for participant recruitment. We plan to conduct data collection soon, and further updates on the recruitment and data collection progress will be provided as the study advances. The AI-powered chatbot offers a novel approach to improving PrEP care utilization among Black cisgender women, with opportunities to reduce barriers to care and facilitate a stigma-free environment. However, challenges remain regarding health equity and the digital divide, emphasizing the need for culturally competent design and robust data privacy protocols. The implications of this study extend beyond PrEP care, presenting a scalable model that can address broader health disparities. PRR1-10.2196/59975.

Bayesian inference of sample-specific coexpression networks.

Gene regulatory networks (GRNs) are effective tools for inferring complex interactions between molecules that regulate biological processes and hence can provide insights into drivers of biological systems. Inferring coexpression networks is a critical element of GRN inference, as the correlation between expression patterns may indicate that genes are coregulated by common factors. However, methods that estimate coexpression networks generally derive an aggregate network representing the mean regulatory properties of the population and so fail to fully capture population heterogeneity. Bayesian optimized networks obtained by assimilating omic data (BONOBO) is a scalable Bayesian model for deriving individual sample-specific coexpression matrices that recognizes variations in molecular interactions across individuals. For each sample, BONOBO assumes a Gaussian distribution on the log-transformed centered gene expression and a conjugate prior distribution on the sample-specific coexpression matrix constructed from all other samples in the data. Combining the sample-specific gene coexpression with the prior distribution, BONOBO yields a closed-form solution for the posterior distribution of the sample-specific coexpression matrices, thus allowing the analysis of large data sets. We demonstrate BONOBO's utility in several contexts, including analyzing gene regulation in yeast transcription factor knockout studies, the prognostic significance of miRNA-mRNA interaction in human breast cancer subtypes, and sex differences in gene regulation within human thyroid tissue. We find that BONOBO outperforms other methods that have been used for sample-specific coexpression network inference and provides insight into individual differences in the drivers of biological processes.

Hybrid organisations embedded in resource-poor contexts: linking organisational capabilities and individual capabilities – a slippery ground?

Purpose This paper aims to contribute to research on hybrid organisations operating within the information and communication technology for development paradigm to foster socio-economic inclusion through the capacity building of marginalised individual youth to enable their entry and participation in the formal labour market. Design/methodology/approach Using a qualitative research design, the authors investigated impact sourcing service providers and their beneficiaries to unpack how hybrid organisations fulfil their social mission and to provide a nuanced understanding of their needed capabilities. Data triangulation through document analysis, participant observation and semi-structured interviews were conducted for the empirical findings’ robustness and in-depth analysis. Findings This paper illuminates how hybrids strived to build beneficiaries' capabilities through empirically grounded approaches of “standardised” and “individualised” skills development, which were strongly connected to the perceived capabilities among the beneficiaries. Organisations that practiced the “individualised” approach imparted technical knowledge and, depending on individual needs, relevant social skills. Critically, the empirical findings call into question the effectiveness of the scalability model, as the results showed that the creation of standardised, low-skilled and low-paid jobs contradicts the idealised self-help status that employment advocates. In addition, the findings underline the central role of constant experimentation, resilience and organisational learning in revamping capabilities of hybrid organisations. Remarkably, compared to organisations committed to the “standardised” technical competency path, the results underscored the difficulties that organisations taking an “individualised" approach face in developing organisational capabilities for their financial sustainability. Further, organisations engaged in standardised” skills development were found to readapt their business model to the economic value and kept the wording of dual mission as a narrative. Originality/value The paper makes a conceptual and empirical contribution bringing together two separately developed literature strands – the organisational capability approach and – the individual capability approach, to enhance a more profound understanding of how both capabilities are connected to each other in the dual-mission orientation of hybrids embedded in resource-poor environments, global value chains, Global South, hybrid organisations, organisational and individual capability building.

PSA-FL-CDM: A Novel Federated Learning-Based Consensus Model for Post-Stroke Assessment.

The rapid development of Internet of Things (IoT) technologies and the potential benefits of employing the vast datasets generated by IoT devices, including wearable sensors and camera systems, has ushered in a new era of opportunities for enhancing smart rehabilitation in various healthcare systems. Maintaining patient privacy is paramount in healthcare while providing smart insights and recommendations. This study proposed the adoption of federated learning to develop a scalable AI model for post-stroke assessment while protecting patients' privacy. This research compares the centralized (PSA-MNMF) model performance with the proposed scalable federated PSA-FL-CDM model for sensor- and camera-based datasets. The computational time indicates that the federated PSA-FL-CDM model significantly reduces the execution time and attains comparable performance while preserving the patient's privacy. Impact Statement-This research introduces groundbreaking contributions to stroke assessment by successfully implementing federated learning for the first time in this domain and applying consensus models in each node. It enables collaborative model training among multiple nodes or clients while ensuring the privacy of raw data. The study explores eight different clustering methods independently on each node, revolutionizing data organization based on similarities in stroke assessment. Additionally, the research applies the centralized PSA-MNMF consensus clustering technique to each client, resulting in more accurate and robust clustering solutions. By utilizing the FedAvg federated learning algorithm strategy, locally trained models are combined to create a global model that captures the collective knowledge of all participants. Comparative performance measurements and computational time analyses are conducted, facilitating a fair evaluation between centralized and federated learning models in stroke assessment. Moreover, the research extends beyond a single type of database by conducting experiments on two distinct datasets, wearable and camera-based, broadening the understanding of the proposed methods across different data modalities. These contributions develop stroke assessment methodologies, enabling efficient collaboration and accurate consensus clustering models and maintaining data privacy.

Large Language Models for Wearable Sensor-Based Human Activity Recognition, Health Monitoring, and Behavioral Modeling: A Survey of Early Trends, Datasets, and Challenges.

The proliferation of wearable technology enables the generation of vast amounts of sensor data, offering significant opportunities for advancements in health monitoring, activity recognition, and personalized medicine. However, the complexity and volume of these data present substantial challenges in data modeling and analysis, which have been addressed with approaches spanning time series modeling to deep learning techniques. The latest frontier in this domain is the adoption of large language models (LLMs), such as GPT-4 and Llama, for data analysis, modeling, understanding, and human behavior monitoring through the lens of wearable sensor data. This survey explores the current trends and challenges in applying LLMs for sensor-based human activity recognition and behavior modeling. We discuss the nature of wearable sensor data, the capabilities and limitations of LLMs in modeling them, and their integration with traditional machine learning techniques. We also identify key challenges, including data quality, computational requirements, interpretability, and privacy concerns. By examining case studies and successful applications, we highlight the potential of LLMs in enhancing the analysis and interpretation of wearable sensor data. Finally, we propose future directions for research, emphasizing the need for improved preprocessing techniques, more efficient and scalable models, and interdisciplinary collaboration. This survey aims to provide a comprehensive overview of the intersection between wearable sensor data and LLMs, offering insights into the current state and future prospects of this emerging field.

The Efficacy of a Cellular Approach to Resilience-Oriented Operation of Distribution Grids Based on Eigenvectors of the Laplacian Matrix

This study addresses the critical question for distribution system operators: how to maintain power supply during an impending disaster that could disrupt the grid? To address this challenge, we propose a cost-efficient cellular model for enhancing the resilience of smart distribution grids. This model prioritizes resilience in the face of natural disasters or other disruptions that could impact service delivery. This approach benefits both grid operators and consumers by ensuring reliable power supply while minimizing energy costs. Furthermore, the model's scalability allows it to be applied to distribution grids of varying sizes. The proposed method utilizes an innovative approach to form optimal cellular network configurations within the grid. As the first step in the formation of cellular topology for the grid, the eigenvectors of the Laplacian matrix of the grid will be used to decide on the optimal configurations. Subsequently, a bi-level mixed-integer linear programming model is proposed to decrease the network costs while simultaneously consider potential power transfer scenarios between the cells and the upstream network during both normal and emergency conditions. The researchers validated the effectiveness of the proposed method through simulations on an IEEE 33-bus test system. The results demonstrate outstanding performance, with a significant increase in the resilience index (around 96%) and a substantial reduction in load-shedding costs (around 80%), making the network considerably more robust.

Virtual Prototyping of the RM Active Grid®: a DEM-MBD Study

Numerical simulation is an established tool to improve knowledge of complex systems and to accelerate the development of new prototypes. In this contribution, the patented Active Grid® made by the company RM RUBBLE MASTER HMH GmbH is examined.In an initial step, representative system parameters (spring properties) are found via a multibody dynamic simulation; at this point, without particles acting on the system components. After that, the virtually defined Active Grid® system is loaded with particles in order to include the interaction effects of various bulk materials acting on the system’s mechanical components. For this purpose, a DEM-MBD co-simulation, extending the discrete element method (DEM) towards multibody dynamics (MBD) simulation, is performed, accounting for the bulk materials via DEM and the interacting system components via MBD. This bi-directional co-simulation enables the consideration of interactions between the bulk materials (the particles) and the driven spring-damper system (the moving system components).With this virtual prototype of the Active Grid®, performance benefits and characteristics in terms of screening efficiency for different material types and varying drive speeds are analysed. The overall aim is to achieve a scalable DEM-MBD model that allows virtual prototyping and, especially, optimisation for future system developments.

Oblique pole-side crash assessment using the six-year-old HBM PIPER

Side impact crashes are one of the most dangerous impact scenarios that a child can suffer. Studies by the National Highway Traffic Safety Administration (NHTSA) have shown that the head and Thorax regions are affected severely. The objective of this work is to perform a numerical evaluation of the oblique pole-side test considering the FMVSS 214-P standard to estimate the Head, Neck, and Thorax injuries for a six-year-old child positioned in the rear seat without any Child Restraint System, two configurations were performed for the oblique pole-side impact: a nearside and a far-side positioning configuration. A six-year-old Human Body Model (HBM) denominated Scalable PIPER Child Model, and the Ford Explorer 2003 were used to perform the test in the LS DYNA® software to assess the biomechanics involved in the crash scenarios. The approach considered a comparative case study with the baseline of the six-year-old child PIPER model to ensure that the positioning adjustment has not affected the mesh quality and interior components for the PIPER child model. The outcomes obtained in case 1 show that the modified PIPER child model has slight outcomes at the shoulder and pelvis zone due to the differences in the body positioning and not by the mesh or the interior interaction between the components. The outcomes obtained in case 2 reflect that the nearside setup obtained the higher measurements for the child occupant. The Ac3ms for Head at nearside test to overcome the Side Criteria established by the Assessment Protocol Child Occupant Protection by Euro NCAP, the kinematics behavior demonstrates the importance of researching children in side crashes to enhance child security, especially in the oblique pole side impact.

Optimizing EV charging stations: a simulation-based approach to performance and grid integration

This study addresses the optimization of electric vehicle (EV) charging stations, focusing on enhancing performance and grid integration through a comprehensive simulation approach. By employing advanced simulation tools in Simulink® and MATLAB®, alongside electrical installation planning with SIMARIS®, we meticulously analyze the charging process, infrastructure requirements, and their implications on the power grid. Our results demonstrate significant improvements in charging station efficiency and reliability, highlighting the effectiveness of our proposed control strategies and harmonic mitigation techniques. Notably, the integration of renewable energy sources emerges as a pivotal factor in reducing operational costs and carbon emissions, furthering the sustainability of EV charging solutions. The research delineates the environmental benefits, emphasizing the reduction of greenhouse gas emissions and enhancement of urban air quality, pivotal in the global shift towards cleaner transportation modes. This work contributes valuable insights into the design and grid integration of EV charging stations, offering a scalable model for future infrastructure development. It serves as a critical resource for engineers, policymakers, and stakeholders in the realm of electric mobility, advocating for a strategic transition to EVs supported by robust and efficient charging infrastructure.

Learning Health System to rapidly improve the implementation of a school physical activity policy

BackgroundLearning Health Systems (LHS) – characterised by cycles of evidence generation and application – are increasingly recognised for their potential to improve public health interventions and optimise health impacts; however there is little evidence of their application in the context of public health practice. Here, we describe how an Australian public health unit applied a LHS approach to successfully improve a model of support for implementation of a school-based physical activity policy.MethodsThis body of work was undertaken in the context of a strong research-practice partnership. Core LHS capabilities included: i) partnerships and stakeholder engagement; ii) workforce development and learning health communities; iii) multi-disciplinary scientific expertise; iv) practice data collection and management system; v) evidence surveillance and synthesis; and vi) governance and organisational processes of decision making. Three cycles of data generation and application were used. Within each cycle, randomised controlled trials conducted in NSW primary schools were used to generate data on the support model’s effectiveness for improving schools’ implementation of a government physical activity policy, its delivery costs, and process measures such as adoption and acceptability. Each type of data were analysed independently, synthesised, and then presented to a multi-disciplinary team of researchers and practitioners, in consult with stakeholders, leading to collaborative decisions for incremental improvements to the support model.ResultsCycle 1 tested the first version of the support model (composed of five implementation strategies targeting identified barriers of policy implementation) and showed the model’s feasibility and efficacy for improving schools’ policy implementation. Data-informed changes were made to enhance impact, including the addition of three implementation strategies to address outstanding barriers. Cycle 2 (now, testing a package of eight implementation strategies) established the model’s effectiveness and cost-effectiveness for improving school's policy implementation. Data-informed changes were made to reduce delivery costs, specifically adapting the costliest strategies to reduce in-person contact from external support personnel. Cycle 3 showed that the adaptations minimised the relative cost of delivery without adversely impacting on the effect.ConclusionsThrough this process, we identified an effective, cost-effective, acceptable and scalable policy implementation support model for service delivery. This provides important information to inform or support LHS approaches for other agencies seeking to optimise the health impact of evidence-based interventions.

The Influence of E-Commerce Platforms on Sustainable Agriculture Practices among Smallholder Farmers in Sub-Saharan Africa

Sustainable agriculture is vital for food security and environmental conservation in Sub-Saharan Africa (SSA), where smallholder farmers encounter challenges like limited resources and market access. E-commerce platforms offer potential solutions by enhancing connectivity and sustainable practices. This review assesses the impact of such platforms on smallholder agriculture in the region, analysing the literature, reports, and case studies from 2010 to 2024 sourced from databases like Google Scholar and Scopus. Inclusion criteria focused on market access, economic outcomes, and sustainability. Findings indicate that e-commerce platforms notably improve market access, leading to better prices and reduced post-harvest losses. They also facilitate information flow, encouraging sustainable practices. However, challenges such as digital literacy and infrastructure deficits persist. E-commerce platforms positively influence sustainable agriculture in SSA by improving market access and supporting sustainable practices. Overcoming barriers through policy interventions and capacity-building is crucial for maximising these benefits. Further research is needed to explore long-term impacts and scalable models.

Deep learning in high-frequency trading: Conceptual challenges and solutions for real-time fraud detection

High-frequency trading (HFT) has transformed financial markets by enabling rapid execution of trades, exploiting market inefficiencies, and optimizing trading strategies. However, this speed and complexity also present significant challenges for real-time fraud detection. Deep learning, a subset of machine learning, offers promising solutions to these challenges through its ability to analyze large volumes of data and uncover intricate patterns. This review explores the conceptual challenges and solutions associated with deploying deep learning for fraud detection in HFT environments. One of the primary challenges in implementing deep learning for HFT fraud detection is the sheer volume and velocity of data. HFT systems generate vast amounts of transactional data in milliseconds, necessitating highly efficient and scalable deep learning models. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are particularly suited for this task due to their ability to process and analyze sequential data efficiently. However, these models require substantial computational resources and sophisticated infrastructure to operate in real time. Another significant challenge is the need for high accuracy and low latency in fraud detection. False positives can lead to unnecessary interventions, while false negatives can result in undetected fraudulent activities. Deep learning models must be fine-tuned to balance these risks, employing techniques such as hyperparameter optimization and ensemble learning to enhance their predictive capabilities. Additionally, integrating real-time anomaly detection methods can help identify suspicious activities promptly, reducing the window of opportunity for fraudsters. Data quality and integrity also pose substantial challenges. HFT environments are susceptible to noise and outliers, which can distort model predictions. Ensuring high-quality data through rigorous preprocessing and anomaly filtering is crucial for the accuracy of deep learning models. Techniques such as data augmentation and normalization can further improve model robustness. To address these challenges, a hybrid approach combining deep learning with traditional statistical methods and rule-based systems can be effective. This approach leverages the strengths of each method, providing a comprehensive fraud detection framework that is both accurate and responsive. Additionally, ongoing model retraining and adaptation to evolving fraud patterns are essential to maintain the effectiveness of the system. In conclusion, while deep learning presents significant opportunities for enhancing real-time fraud detection in high-frequency trading, it also requires addressing challenges related to data volume, computational demands, accuracy, and data quality. By employing a hybrid approach and continually refining models, financial institutions can effectively mitigate fraud risks and safeguard their trading operations.

Fine-grained personalized federated learning via transformer in the transformer framework

Federated learning (FL) stands out as one of the most decentralized and privacy-preserving machine learning techniques. FL enables collaborative model training across multiple clients while preserving the privacy of their data. An extension of this method, namely, personalized federated learning (PFL), seeks to address client data heterogeneity by tailoring individualized models. Existing PFL methods that use transformers do not always take into account how federated averaging algorithms (a common FL algorithm that takes a weighted average of model parameters) can aggregate the self-attention mechanism. This makes it harder for transformers to work in federated settings. In addition, these approaches typically underperform on fine-grained tasks. To overcome these limitations, this study develops a novel transformer-in-transformer facilitated FL framework that not only aggregates global model parameters but also pinpoints personalized parameters for each client. This framework departs from traditional methods by implementing an “adaptive-personalization” mechanism, which not only enhances inter-client cooperation but also improves the model's scalability and adaptability. On the server side, a hypernetwork is developed to generate personalized projection matrices for use in the self-attention layers. These matrices determine the query, key, and value parameters for each client. Experimental results demonstrate that the model equipped with this adaptive-personalization mechanism considerably outperforms traditional models in scenarios with heterogeneous data, thus validating its effectiveness and efficiency.

Assessing the spatial transferability of calibration models across a low-cost sensors network

Low-cost sensor networks (LCSNs) are expanding worldwide to gather high spatiotemporal resolution data due to their economic feasibility and compact size. The reliability of LCS-recorded data is limited due to their calibration dependencies in the field. Previous studies have focused on the development of LCS calibration models by co-location with the regulatory monitoring stations. However, it is challenging to calibrate LCS in the field for countries with limited infrastructure for air quality monitoring, pointing towards the need for transferable calibration models. Only a few studies have addressed this challenge and provide no information on the factors that may affect the performance of transferable calibration models. Here, we examined the spatial transferability of the calibration models developed using machine learning (ML) algorithms for an LCSN with twenty-two (22) sites in NCT-Delhi. The site-specific calibration models performed well at each site with high R2 and significantly low RMSE values. These models were transferred to the other sites, and the effect of distance between the sites (D), source composition, PM ratios, and particle size distribution (PSD) on the transferability of calibration models was investigated. The models developed at the Mundka (S10) and Punjabi Bagh (S16) sites complied with the evaluation criterion (R2 ≥ 0.70) for each site, irrespective of the distance between the sites. Furthermore, PM ratios reported by the LCSs did not significantly differ across sites, suggesting that the PMS algorithm provides a proxy of the size-resolved mass fractions. Evaluation of the PSD at different sites supported our findings. We also introduced the concept of selecting representative locations for LCS co-location by computing transferability scores using k-means clustering and presented a reference map for NCT-Delhi for developing scalable calibration models.