检索结果-内蒙古大学图书馆

computers in Biology and Medicine 2025年 189卷 110011-110011页

作者： Saad, Eman Kishk, Sherif Ali-Eldin, Amr Saleh, Ahmed I. Communication and Information Technology Center Mansoura University Egypt Electronics and Communication Engineering Dept. Mansoura University Egypt Computer Engineering & Control Systems Dept. Mansoura University Egypt

Drug-Drug Interactions (DDI) and Chemical-Protein Interactions (CPI) detection are crucial for patient safety, as unidentified interactions may lead to severe Adverse Drug Reactions (ADRs). While extensive DDI and CPI information exists within biomedical literature, manual extraction by experts proves time-intensive and resource-demanding. This study presents SB-AGT: Stochastic Beam Search-enhanced Attention-based Gumbel Tree an innovative approach to automated DDI and CPI extraction through an attention-based architecture incorporating a modified Gumbel-Tree method with stochastic beam search optimization. SB-AGT leverages multi-head attention mechanisms and enhanced latent tree structures to effectively capture complex syntactic features and drug relationship patterns. Also, the methodology includes comprehensive preprocessing protocols and addresses dataset imbalance through hybrid sampling techniques. Experimental validation on the DDIExtraction 2013 and CHEMPROT datasets demonstrates the model's effectiveness, achieving precision, surpassing traditional approaches and showing competitive performance compared to contemporary pre-trained models. Through parametric analysis, we establish optimal beam search configurations that maximize the model's extraction accuracy. © 2025 Elsevier Ltd

关键词： Stochastic models

Enhancing Train Transportation in Sri Lanka: A Smart IOT based Multi-Subsystem Approach using MQTT

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Balasingham, Dhanushka Samarathunga, Sadeesha Bandara, Anuththara Arachchige, Gayantha Godakanda Gamage, Narmada Wijayaraja, Jaliya L. Dept. of Computer Systems Engineering Sri Lanka Institute of Information Technology Sri Lanka

This research proposes a system as a solution for the challenges faced by Sri Lanka’s historic railway system, such as scheduling delays, overcrowding, manual ticketing, and management inefficiencies. It proposes a multi-subsystem approach, incorporating GPS tracking, RFID-based e-ticketing, seat reservation, and vision-based people counting. The GPS-based real-time train tracking system performs accurately within 24 meters, with the MQTT protocol showing twice the speed of the HTTP-based system. All subsystems use the MQTT protocol to enhance efficiency, reliability, and passenger experience. The study’s data and methodology demonstrate the effectiveness of these innovations in improving scheduling, passenger flow, and overall system performance, offering promising solutions for modernizing Sri Lanka’s railway infrastructure. © 2025, CC BY.

关键词： Global positioning system

Predicting Lifespan of Ground-to-Air Multipath Components in mmWave UAV Channels

学校读者我要写书评

暂无评论

Predicting Lifespan of Ground-to-Air Multipath Components in...

IEEE Conference on Wireless Communications and Networking

作者： Wahab Khawaja Rune H. Jacobsen Sajid Hussain Ismail Guvenc Dept. Electrical and Computer Engineering Aarhus University Aarhus Denmark Dept. Computer Systems Engineering Mirpur Univ. Science and Technology Mirpur AJK Pakistan Dept. Electrical and Computer Engineering North Carolina State University Raleigh NC USA

ISBN: (数字)9798350368369

ISBN: (纸本)9798350368376

In mobile ground-to-air (GA) propagation channels, the birth and death of multipath components (MPCs) are frequently observed, and the wide-sense stationary uncorrelated scattering (WSSUS) assumption does not always hold. Several methods exist for tracking the birth and death of MPCs, however, to the best of knowledge of authors, there is no existing literature that addresses the prediction of the lifespan of the MPCs in non-WSSUS GA propagation channels. In this work, we consider the GA channel as non-WSSUS and individual MPCs across receiver positions are represented as time series based on the Euclidean distance between channel parameters of the MPCs. These time series representations, referred to as path bins, are analyzed using a semi-Markov chain model. The channel parameter variations and dependencies between path bins are used to predict the lifespan of path bins using weighted sum method, machine learning classifiers, and deep neural networks. For comparison, the birth and death of path bins are also modeled using a Poisson distribution and a Markov chain. Simulation results demonstrate that deep neural networks offer highly accurate predictions for the lifespan (including death) of MPC path bins in the considered GA propagation scenario.

关键词： Analytical models Simulation Time series analysis Scattering Artificial neural networks Machine learning Receivers Euclidean distance Autonomous aerial vehicles Millimeter wave communication

WQuatNet: Wide range quaternion-based head pose estimation

学校读者我要写书评

暂无评论

Journal of King Saud University - computer and Information Sciences 2025年第3期37卷 1-14页

作者： Algabri, Redhwan Shin, Hyunsoo Abdu, Ahmed Bae, Ji-Hun Lee, Sungon Dept. of Computer Science & Engineering College of AI Convergence Sejong University Seoul 05006 South Korea Research Institute of Engineering and Technology Hanyang University Ansan 15588 South Korea School of Software Northwestern Polytechnical University Xi’an 710072 China Department of Robotics Hanyang University Ansan 15588 South Korea

Head pose estimation (HPE) is a critical task for numerous applications ranging from human-computer interaction, healthcare, and robotics, to surveillance. Most existing methods employ Euler angles as a representation, which often face challenges such as a gimbal lock, especially in full-range rotation scenarios or rotation matrices that require nine parameters. This study introduces WQuatNet, a novel deep learning-based model that leverages the quaternion representation, which uses only four parameters, to avoid this challenge. WQuatNet was designed based on a landmark-free HPE method to predict head poses across the full-range angles of 360∘ from images. Landmark-free methods bypass the need for explicit detection of facial landmarks;instead, they leverage the entire image to estimate the head orientation. The model incorporates a RepVGG-D2se backbone for robust feature extraction and introduces two loss functions tailored for quaternion predictions. Our experimental results on multiple HPE datasets covering both narrow- and full-range angles demonstrate that WQuatNet outperforms the state-of-the-art (SOTA) approaches in terms of accuracy. The performance of the proposed HPE was evaluated using the CMU, AGORA, BIWI, AFLW2000, and 300W-LP datasets. We also perform ablation studies and error analyses to validate the significance of each component of the model. © The Author(s) 2025.

关键词： Deep neural network Full range of rotation Head pose estimation Quaternion

Predicting Lifespan of Ground-to-Air Multipath Components in mmWave UAV Channels

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Khawaja, Wahab Jacobsen, Rune H. Hussain, Sajid Guvenc, Ismail Dept. Electrical and Computer Engineering Aarhus University Aarhus8000 Denmark Dept. Computer Systems Engineering Mirpur Univ. Science and Technology AJK Mirpur10250 Pakistan

关键词： Markov chains

Dynamic Network Optimization with Deep Learning and Deployment of Intelligent Reflecting Surfaces

学校读者我要写书评

暂无评论

Dynamic Network Optimization with Deep Learning and Deployme...

International Conference on Ubiquitous Information Management and Communication (IMCOM)

作者： Nayeon Jang Jeonghoon Bae Tao Tianyu Gyurin Byun Phuoc-Nguyen Bui Tae-Jin Lee Hyunseung Choo Dept. of AI Systems Engineering Sungkyunkwan University Suwon South Korea Dept. of Electrical and Computer Engineering Sungkyunkwan University Suwon South Korea

ISBN: (数字)9798331507817

ISBN: (纸本)9798331507824

The proliferation of 5G technologies has presented challenges such as signal attenuation and high energy consumption. This study proposes a novel approach to enhance wireless networks by integrating Intelligent Reflecting Surfaces (IRS) and Artificial Intelligence (AI) to optimize signal propagation and predict the Received Signal Strength Indicator (RSSI). A new framework combines K-means clustering for IRS deployment with a Temporal Convolutional Network-Gated Recurrent Unit (TCN-GRU) model for signal prediction. Using data from 15,000 Wi-Fi access points in Herefordshire, UK, the results demonstrate significant improvements in signal strength, with the proposed model achieving a Mean Squared Error (MSE) of 37% lower than the existing models, indicating higher prediction accuracy. In addition to enhancing signal quality, the framework achieves throughput increases of 21.26% with single-group IRS and 64.91% with multi-group IRS. These results validate the proposed the potential of the framework to improve performance, making it suitable for next-generation wireless networks.

关键词： Deep learning Wireless networks Predictive models Throughput Data models Received signal strength indicator Artificial intelligence Optimization Next generation networking Wireless fidelity

Advanced Tool Learning and Selection System (ATLASS): A Closed-Loop Framework Using LLM

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Haque, Mohd Ariful Williams, Justin Siddique, Sunzida Islam, Md. Hujaifa Ali, Hasmot Gupta, Kishor Datta George, Roy Dept. of Cyber Physical Systems Clark Atlanta University United States Dept. of Computer Science and Engineering Daffodil International University Bangladesh Dept. of Computer Science and Engineering Ahsanullah University of Science and Technology Bangladesh

The combination of LLM agents with external tools enables models to solve complex tasks beyond their knowledge base. Human-designed tools are inflexible and restricted to solutions within the scope of pre-existing tools created by experts. To address this problem, we propose ATLASS, an advanced tool learning and selection system designed as a closed-loop framework. It enables the LLM to solve problems by dynamically generating external tools on demand. In this framework, agents play a crucial role in orchestrating tool selection, execution, and refinement, ensuring adaptive problem-solving capabilities. The operation of ATLASS follows three phases: The first phase, Understanding Tool Requirements, involves the Agents determining whether tools are required and specifying their functionality;the second phase, Tool Retrieval/Generation, involves the Agents retrieving or generating tools based on their availability;and the third phase, Task Solving, involves combining all the component tools necessary to complete the initial task. The Tool Dataset stores the generated tools, ensuring reusability and minimizing inference cost. Current LLM-based tool generation systems have difficulty creating complex tools that need APIs or external packages. In ATLASS, we solve the problem by automatically setting up the environment, fetching relevant API documentation online, and using a Python interpreter to create a reliable, versatile tool that works in a wider range of situations. OpenAI GPT-4.0 is used as the LLM agent, and safety and ethical concerns are handled through human feedback before executing generated code. By addressing the limitations of predefined toolsets and enhancing adaptability, ATLASS serves as a real-world solution that empowers users with dynamically generated tools for complex problem-solving. © 2025, CC BY-SA.

关键词： Reusability

Recognition of American sign language using modified deep residual CNN with modified canny edge segmentation

学校读者我要写书评

暂无评论

Multimedia Tools and Applications 2025年 1-28页

作者： Hariharan, U. Devarajan, Devikanniga Kumar, P. Santhosh Rajkumar, K. Meena, M. Akilan, T. Department of Computer Science and Engineering Apex Institute of Technology Chandigarh University Punjab Mohali140413 India Department of Computer Science and Engineering GITAM University Karnataka Bengaluru561203 India Department of Information Technology Faculty of Engineering and technology SRM Institute of Science and Technology Ramapuram Tamil Nadu Chennai600089 India Department of Data Science and Business Systems School of Computing Faculty of Engineering and Technology SRM Institute of Science and Technology Kattankulathur Tamil Nadu Chennai603203 India Department of Computer Science and Engineering Sri Sairam Engineering College Tamil Nadu Tambaram602109 India Research Coordinator Dept. of IQAC Amity University Kolkata West Bengal 700135 India

American Sign Language (ASL) recognition aims to recognize hand gestures, and it is a crucial solution to communicating between the deaf community and hearing people. However, existing sign language recognition algorithms still have some drawbacks, such as difficulty recognizing hand movements and low recognition accuracy for most sign language recognition. A Modified Convolutional Neural Network (MCNN) deep residual 101 classifier-based American Sign Language identification system has been developed to address this problem. There are three main parts presents in the method. The first part is pre-processing the images to remove the noise, enhance the contrast of the picture, adjust the contrast level and smoothen the picture using various filters. The second part is the segmentation, and it’s used to partition the image using a modified canny edge detection method by removing all weak edges present in the image. Finally, classification will be done using the Modified CNN deep residual 101 classifiers. By classifying the image, the American Sign Language is accurately identified. This process is conducted through images. The outcome shows that the suggested approach has a 0.97% accuracy and a 0.05% False Positive Rate. Other CNN architectures such as resNet 50 and resNet 18 reached 0.90% and 0.80% accuracy, respectively, which is lower than our proposed method. In addition, the resNet 101 classifier effectively recognizes the difficult hand gestures through the image data. It obtains high recognition accuracy for 36 signs is 0 to 9 numbers and a to z alphabets from American Sign Language. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2025.

关键词： Convolutional neural networks

Deep Reinforcement Learning Algorithms for Option Hedging

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Neagu, Andrei Godin, Frédéric Kosseim, Leila Dept. of Computer Science and Software Engineering Concordia University Montréal Canada Dept. of Mathematics & Statistics Concordia University Montréal Canada

Dynamic hedging is a financial strategy that consists in periodically transacting one or multiple financial assets to offset the risk associated with a correlated liability. Deep Reinforcement Learning (DRL) algorithms have been used to find optimal solutions to dynamic hedging problems by framing them as sequential decision-making problems. However, most previous work assesses the performance of only one or two DRL algorithms, making an objective comparison across algorithms difficult. In this paper, we compare the performance of eight DRL algorithms in the context of dynamic hedging;Monte Carlo Policy Gradient (MCPG), Proximal Policy Optimization (PPO), along with four variants of Deep Q-Learning (DQL) and two variants of Deep Deterministic Policy Gradient (DDPG). Two of these variants represent a novel application to the task of dynamic hedging. In our experiments, we use the Black-Scholes delta hedge as a baseline and simulate the dataset using a GJR-GARCH(1,1) model. Results show that MCPG, followed by PPO, obtain the best performance in terms of the root semi-quadratic penalty. Moreover, MCPG is the only algorithm to outperform the Black-Scholes delta hedge baseline with the allotted computational budget, possibly due to the sparsity of rewards in our environment. © 2025, CC BY.

关键词： Learning algorithms

Dynamic SpikFormer: Low-Latency & Energy-Efficient Spiking Neural Networks with Dynamic Time Steps for Vision Transformers

学校读者我要写书评

暂无评论

Dynamic SpikFormer: Low-Latency & Energy-Efficient Spiking N...

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Gourav Datta Zeyu Liu Anni Li Peter A. Beerel Dept. of Electrical Computer & Systems Engineering Case Western Reserve University Cleveland USA Ming Hsieh Dept. of Electrical and Computer Engineering University of Southern California Los Angeles USA

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Spiking Neural Networks (SNNs) have emerged as a popular spatio-temporal computing paradigm for complex vision tasks. Recently proposed SNN training algorithms have significantly reduced the number of time steps (down to 1) for improved latency and energy efficiency, however, they target only convolutional neural networks (CNN). These algorithms, when applied to the recently spotlighted vision transformers (ViT), either require a large number of time steps or fail to converge. Based on the analysis of the histograms of the ANN and SNN activation maps, we hypothesize that each ViT block has a different sensitivity to the number of time steps. We propose a novel training framework that dynamically allocates the number of time steps to each ViT module depending on a trainable score assigned to each timestep. In particular, we generate a scalar binary time step mask that filters spikes emitted by each neuron in a leaky-integrate-and-fire (LIF) layer. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), except for the input embedding layer, in contrast to expensive multiply-and-accumulates (MAC) needed in traditional ViTs. This yields significant improvements in energy efficiency. We evaluate our training framework and resulting SNNs on image recognition tasks including CIFAR10, CIFAR100, and ImageNet with different ViT architectures. We obtain a test accuracy of 95.97% with 4.97 time steps with direct encoding on CIFAR10.

关键词： Training computer vision Sensitivity Neurons Signal processing algorithms Spiking neural networks Transformers Energy efficiency Convolutional neural networks Speech processing