Anomaly Detection (AD) play a significant part in risk management and data security among various domains like medical image recognition, economic insurance security, Internet of Things (IoT) management. However, seve...
详细信息
Peer-to-peer learning is an increasingly popular framework that enables beyond-5G distributed edge devices to collaboratively train deep neural networks in a privacy-preserving manner without the aid of a central serv...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Peer-to-peer learning is an increasingly popular framework that enables beyond-5G distributed edge devices to collaboratively train deep neural networks in a privacy-preserving manner without the aid of a central server. Neural network training algorithms for emerging environments, e.g., smart cities, have many design considerations that are difficult to tune in deployment settings – such as neural network architectures and hyperparameters. This presents a critical need for characterizing the training dynamics of distributed optimization algorithms used to train highly nonconvex neural networks in peer-to-peer learning environments. In this work, we provide an explicit characterization of the learning dynamics of wide neural networks trained using popular distributed gradient descent (DGD) algorithms. Our results leverage both recent advancements in neural tangent kernel (NTK) theory and extensive previous work on distributed learning and consensus. We validate our analytical results by accurately predicting the parameter and error dynamics of wide neural networks trained for classification tasks.
The paper introduces an event chronology model for Hindu mythological texts, with a focus on the Ramayana's Bala-kanda. It captures and structures sentence-level events based on their chronological order to better...
详细信息
ISBN:
(数字)9798331520762
ISBN:
(纸本)9798331520779
The paper introduces an event chronology model for Hindu mythological texts, with a focus on the Ramayana's Bala-kanda. It captures and structures sentence-level events based on their chronological order to better understand narrative progression and character interactions. A manually crafted dataset of 1217 events, involving 214 characters across 75 chapters, is used to generate a knowledge graph. The graph organizes events by sequence, linking characters, actions, and chapters within the storyline. The event chronology is mapped by establishing relationships between events, ensuring they are connected in the proper sequence to reflect the narrative flow. This approach enables a deeper understanding of how each event contributes to the overarching story, providing insights into both the individual actions of characters and their significance within the broader mythological narrative. Transformer-based models (BERT, DistilBERT, and RoBERTa) were utilized for predicting subsequent events, with RoBERTa showing the highest accuracy in preserving the correct event chronology. This structure is key to analyzing the sequential development of the Ramayana's storyline.
Vendor selection in supply chain management is a complex but extremely important process that requires the evaluation of multiple factors such as cost, quality, delivery time, and responsiveness. This study presents a...
详细信息
ISBN:
(数字)9798331531935
ISBN:
(纸本)9798331531942
Vendor selection in supply chain management is a complex but extremely important process that requires the evaluation of multiple factors such as cost, quality, delivery time, and responsiveness. This study presents a new hybrid methodology that makes the vendor selection process more efficient by combining tabu search optimization and neural network analysis. The proposed model operates in two stages. In the first stage, tabu search algorithms systematically analyze the solution space and identify optimal vendor combinations based on predefined criteria. In the second stage, these selected combinations are evaluated through a neural network. This trained network has the ability to recognize patterns in performance indicators such as cost efficiency, delivery time, and operational wastage. Tests conducted on simulated data clearly demonstrated the superior performance of this hybrid model compared to traditional single-method approaches. The model significantly improved key parameters such as Mean Absolute Error and R 2 score. Moreover, it also proved to be adaptable to changing conditions. This framework not only addresses challenges such as scalability and integration of qualitative-quantitative factors in vendor selection, but also highlights the effectiveness of combining optimization algorithms and machine learning techniques for complex supply chain decisions. This model not only serves as a practical decision-support tool for procurement managers but also lays a strong foundation for the development of hybrid AI techniques in the field of operational optimization.
—Phishing URL detection is crucial in cybersecurity as malicious websites disguise themselves to steal sensitive information. Traditional machine learning techniques struggle to perform well in complex real-world sce...
详细信息
Traffic sign detection in foggy weather is a challenging task for unmanned driving systems. Fog introduces noise pollution to the images, which will affect the detection accuracy. In this paper, a traffic sign detecti...
详细信息
This study presents a comparative analysis of ten pre-trained convolutional neural network (CNN) models, evaluated across three remote sensing datasets: EuroSat, NWPU, and Earth Hazards (Land Sliding). We investigate ...
详细信息
A novel personalized tour recommendation model, the Happiness Model (HM), is presented. The HM optimizes itineraries by considering traveler satisfaction as a function of time and maximizing it over the trip duration....
详细信息
A novel personalized tour recommendation model, the Happiness Model (HM), is presented. The HM optimizes itineraries by considering traveler satisfaction as a function of time and maximizing it over the trip duration. The model integrates the Item Constraints Data Model (ICDM) to reduce data dimensionality and search space. By considering various activities within different points of interest (POIs) and minimizing wasted time, the HM overcomes the limitations of existing methods. Unlike existing POI-centric models, the HM is time-centric, creating tour recommendations that maximize user satisfaction throughout the trip. Experimental results demonstrate the model’s effectiveness in generating personalized tour recommendations aligned with user preferences. The HM achieves an average satisfaction score of 0.85 across multiple datasets, outperforming traditional models such as the Time-Dependent Orienteering Problem with Time Windows (TOPTW), which achieves an average score of 0.72. Additionally, the HM reduces waiting times by 30% and increases the number of recommended POIs by 20% compared to existing methods. These results highlight the HM’s ability to provide more efficient and enjoyable travel experiences. Copyright 2025 Alatiyyah Distributed under Creative Commons CC-BY 4.0
Dementia, a progressive neurodegenerative disorder, affects memory, reasoning, and daily functioning, creating challenges for individuals and healthcare systems. Early detection is crucial for timely interventions tha...
详细信息
Electric vehicles are rapidly gaining popularity as a sustainable alternative to conventional gasoline. In urban areas, chargers with different ratings can accommodate the diverse needs of electric vehicles. However, ...
详细信息
暂无评论