检索结果-内蒙古大学图书馆

Microservice Deployment for Satellite Edge AI Inference via Deep Reinforcement Learning

学校读者我要写书评

暂无评论

Microservice Deployment for Satellite Edge AI Inference via ...

IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC)

作者： Zhiyong Yu Hei Victor Cheng Zhanpeng Yang Xin Liu Yuning Jiang Yong Zhou Yuanming Shi School of Information Science and Technology ShanghaiTech University Shanghai China Department of Electrical and Computer Engineering Aarhus University Denmark Automatic Control Laboratory École Polytechnique Fédérale de Lausanne (EPFL) Switzerland

ISBN: (数字)9798350362244

ISBN: (纸本)9798350362251

Artificial intelligence (AI) is critical in evolving 5G and developing 6G networks, running on edge devices, and solving resource management challenges. The burgeoning number of edge devices draws attention to the potential of low-earth orbit (LEO) satellite networks with their onboard computing capabilities for edge inference. This paper explores LEO scenarios where multiple remote sensing edge AI inference tasks concurrently process data from a single source. However, due to there being parts with the same functions between different AI applications, traditional monolithic edge AI architecture must be deployed repeatedly and falls short in efficiently harnessing the heterogeneous resources of LEO satellite networks. To solve this problem, we utilize the microservice architecture to decouple a single AI application into several independent microservices to reuse these same functions. However, due to the high latency caused by multiple microservices’ communication, we need to design a deployment strategy to fully utilize resources to reduce the service latency. We present a microservice deployment model to minimize the total service latency across all AI applications and meet resource constraints with the constraints of hardware, energy, and memory limitations. This latency optimization problem is rewritten as a Markov decision process (MDP) to effectively deal with the challenge posed by the time-varying transmission rate caused by satellite mobility. To increase the training data utilization, we employ a Proximal Policy Optimization (PPO) based reinforcement learning algorithm to meet the dynamic environment challenge. Finally, we obtain a sub-optimal solution with minimal accuracy loss and an acceptable solution time.

关键词： Satellites Accuracy Heuristic algorithms Satellite broadcasting Microservice architectures Low earth orbit satellites Edge AI computer architecture Resource management Optimization

SIVD: Dataset of Iranian Vehicles for Real-Time Multi-Camera Video Tracking and Recognition

学校读者我要写书评

暂无评论

SIVD: Dataset of Iranian Vehicles for Real-Time Multi-Camera...

International Conference of Signal Processing and Intelligent systems (ICSPIS)

作者： Farbod Siahkali Seyed Amirmahdi Alavi Mehdi Tale Masouleh Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran Human and Robot Interaction Laboratory School of Computer Science Islamic Azad University Central Tehran Branch Tehran Iran

ISBN: (纸本)9781665476249

In this paper, a new publicly available 1 web-Scraped Iranian Vehicle Dataset (SIVD) for simultaneous real-time vehicle tracking and recognition is proposed. The datasets provided for Iranian cars in the literature have two fundamental problems. First, the lack of images from different angles, and second, the small number of classes compared to the dispersion of car models in the real world. Therefore, for the purposes of this paper, Iranian vehicle images from car sales websites are collected, and the SIVD dataset is proposed which contains 29 classes and 36,705 images. This paper aims at developing a classification network for Iranian vehicle recognition and implement a real-time tracking system using the YOLOv5 network to perform real-time vehicle model recognition and tracking tasks simultaneously. The ResNet50 achieved an accuracy of 99.29%, the highest among the investigated classification networks. 1 https://***/Farbod-Siahkali/Iranian-Vehicle-Tracking-and-Recognition

关键词： Representation learning Image recognition Video tracking Machine learning algorithms Neural networks Real-time systems Automobiles

Block-Wise Index Modulation and Receiver Design for High-Mobility OTFS Communications

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Qian, Mi Ji, Fei Ge, Yao Wen, Miaowen Cheng, Xiang Poor, H. Vincent The School of Electronic and Information Engineering South China University of Technology Guangzhou510641 China The Continental-NTU Corporate Lab Nanyang Technological University Singapore The State Key Laboratory of Advanced Optical Communication Systems and Networks Department of Electronics School of Electronics Engineering and Computer Science Peking University Beijing100871 China The Department of Electrical and Computer Engineering Princeton University PrincetonNJ08544 United States

As a promising technique for high-mobility wireless communications, orthogonal time frequency space (OTFS) has been proved to enjoy excellent advantages with respect to traditional orthogonal frequency division multiplexing (OFDM). Although multiple studies have considered index modulation (IM) based OTFS (IM-OTFS) schemes to further improve system performance, a challenging and open problem is the development of effective IM schemes and efficient receivers for practical OTFS systems that must operate in the presence of channel delays and Doppler shifts. In this paper, we propose two novel block-wise IM schemes for OTFS systems, named delay-IM with OTFS (DeIM-OTFS) and Doppler-IM with OTFS (DoIM-OTFS), where a block of delay/Doppler resource bins are activated simultaneously. Based on a maximum likelihood (ML) detector, we analyze upper bounds on the average bit error rates for the proposed DeIM-OTFS and DoIM-OTFS schemes, and verify their performance advantages over the existing IM-OTFS systems. We also develop a multi-layer joint symbol and activation pattern detection (MLJSAPD) algorithm and a customized message passing detection (CMPD) algorithm for our proposed DeIM-OTFS and DoIM-OTFS systems with low complexity. Simulation results demonstrate that our proposed MLJSAPD and CMPD algorithms can achieve desired performance with robustness to the imperfect channel state information (CSI). © 2023, CC BY.

关键词： Message passing

Class Incremental Learning for Knowledge Acquisition Based on Prototype Network Gating Mechanism

学校读者我要写书评

暂无评论

Class Incremental Learning for Knowledge Acquisition Based o...

Medical Artificial Intelligence (MedAI), IEEE International Conference on

作者： Yuming Feng Xiaolong Jin Haozhe Liang Jiafeng Guo School of Computer Science and Technology University of Chinese Academy of Sciences CAS Key Laboratory of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences Institute of Systems Engineering Academy of Military Science

ISBN: (数字)9798350377613

ISBN: (纸本)9798350377620

With the development of information technology, the medical field has accumulated a large amount of medical data, which continues to grow. Efficient extraction and effective utilization of knowledge from this vast amount of medical data are crucial for realizing its immense value. Traditional knowledge acquisition algorithms face challenges in effectively identifying and acquiring new types of knowledge in practical applications. To address this, this paper proposes a class-incremental learning method based on a prototype network gating mechanism, called PGMoE. This method integrates multiple experts through a prototype network-based gating mechanism, dynamically adjusting the model's response to new and old knowledge, thereby effectively mitigating the issue of catastrophic forgetting in class-incremental learning. PGMoE not only improves the stability and accuracy of the model in continuous learning processes but also ensures the model's ability to remember historical knowledge. Comparative experiments on multiple datasets with existing class-incremental learning methods demonstrate that PGMoE has significant performance advantages in handling continuous learning tasks.

关键词： Learning systems Accuracy Incremental learning Knowledge acquisition Scalability Heuristic algorithms Prototypes Logic gates Stability analysis Information technology

Optimizing Mobile-Edge AI-Generated Everything (AIGX) Services by Prompt engineering: Fundamental, Framework, and Case Study

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Liu, Yinqiu Du, Hongyang Niyato, Dusit Kang, Jiawen Cui, Shuguang Shen, Xuemin Zhang, Ping The School of Computer Science and Engineering Nanyang Technological University Singapore The School of Automation Guangdong University of Technology China The Chinese University of Hong Kong Shenzhen China The Department of Electrical and Computer Engineering University of Waterloo Canada The State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications China

As the next-generation paradigm for content creation, AI-Generated Content (AIGC), i.e., generating content automatically by Generative AI (GAI) based on user prompts, has gained great attention and success recently. With the ever-increasing power of GAI, especially the emergence of Pretrained Foundation Models (PFMs) that contain billions of parameters and prompt engineering methods (i.e., finding the best prompts for the given task), the application range of AIGC is rapidly expanding, covering various forms of information for human, systems, and networks, such as network designs, channel coding, and optimization solutions. In this article, we present the concept of mobile-edge AI-Generated Everything (AIGX). Specifically, we first review the building blocks of AIGX, the evolution from AIGC to AIGX, as well as practical AIGX applications. Then, we present a unified mobile-edge AIGX framework, which employs edge devices to provide PFM-empowered AIGX services and optimizes such services via prompt engineering. More importantly, we demonstrate that suboptimal prompts lead to poor generation quality, which adversely affects user satisfaction, edge network performance, and resource utilization. Accordingly, we conduct a case study, showcasing how to train an effective prompt optimizer using ChatGPT and investigating how much improvement is possible with prompt engineering in terms of user experience, quality of generation, and network performance. © 2023, CC BY.

关键词： Network performance

Reversible Data Hiding in Encrypted Image via Joint Encoding of Multiple MSB and Pixel Difference

学校读者我要写书评

暂无评论

Security and Communication Networks 2023年第1期2023卷

作者： Kong, Ping Fu, Di Huang, Lin Zhou, Liang Li, Jian Qin, Chuan Shanghai Key Laboratory of Molecular Imaging Jiading District Central Hospital Affiliated Shanghai University of Medicine and Health Sciences Shanghai201318 China School of Medical Instrument and Food Engineering University of Shanghai for Science and Technology Shanghai200093 China School of Optical-Electrical and Computer Engineering University of Shanghai for Science and Technology Shanghai200093 China Shandong Provincial Key Laboratory of Computer Networks Jinan250353 China

Reversible data hiding in encrypted image (RDHEI) has become a research hotspot, which can effectively protect image content privacy. An RDHEI scheme based on the joint encoding of multiple MSBs (most significant bits) and pixel difference is proposed in this paper. A block-based image encryption method is adopted on the content owner side, which can securely protect the image contents while retaining the spatial correlation within each block. By the joint encoding strategy of multiple MSB and pixel difference, the redundancy within the bit plane is sufficiently compressed to accommodate more additional data;thus, a high embedding rate can be achieved. According to different kinds of available keys, image decryption and data extraction can be separably conducted on the receiver side. Experimental results show that our scheme can achieve a higher embedding rate than some state-of-the-art schemes. © 2023 Ping Kong et al.

关键词： Pixels

Modular Multilevel DC-DC Converter With Inherent Bipolar Operation Capability for Resilient Bipolar MVDC Grids

学校读者我要写书评

暂无评论

CPSS Transactions on Power Electronics and Applications 2022年第1期7卷 37-48页

作者： Shanshan Zhao Yu Chen Shenghui Cui Jingxin Hu State Key Laboratory of Advanced Electromagnetic Engineering and Technology School of Electrical and Electronic Engineering Huazhong University of Science and Technology China Department of Electrical and Computer Engineering Seoul National University Korea Institute for Power Generation and Storage Systems E. ON Energy Research Center RWTH Aachen University Germany

Bipolar medium-voltage dc (MVDC) distribution systems are of great interests nowadays due to its high availability and reliability. A bipolar MVDC grid can be established by the installation of power-balancers or directly by converters with inherent bipolar operation capability to avoid the extra cost and volume. Therefore, this paper proposes a modular multilevel dc-dc converter (MMDC) with inherent bipolar operation capability for the interconnection of bipolar MVDC grids and LVDC girds. Considering that the MMDC is normally only capable of operating with monopole MVDC distribution systems due to the symmetric structure and operation scheme, a center-tapped high-frequency interface transformer is employed on the MV side of the MMDC. Based on the concept of the flux dc-bias cancellation, a dedicated operation method is proposed, whereby the power flows of the two MVDC poles can be regulated independently with a simple control scheme, and no penalty of increased current rating is imposed on the semiconductor devices or the interface transformer compared to the conventional MMDC with monopole operation. Additionally, the MMDC can realize single-pole operation in case one pole is faulty and deliver at least 50% of the rated power capacity, which significantly enhance the reliability and availability of the power delivery. The validity of the proposed bipolar operation scheme of the MMDC has been verified by both simulations and experiments with a down-scale prototype.

关键词： Multilevel converters Semiconductor devices Prototypes DC-DC power converters Transformers Valves Power electronics

Green Routing Game: Strategic Logistical Planning using Mixed Fleets of ICEVs and EVs

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Sasahara, Hampei Dán, György Amin, Saurabh Sandberg, Henrik The Department of Systems and Control Engineering School of Engineering Tokyo Institute of Technology Tokyo152-8552 Japan The Division of Network and Systems Engineering School of Electrical Engineering and Computer Science Kth Royal Institute of Technology StockholmSE-100 44 Sweden The Laboratory for Information and Decision Systems Massachusetts Institute of Technology CambridgeMA02139 United States The Division of Decision and Control Systems School of Electrical Engineering and Computer Science Kth Royal Institute of Technology StockholmSE-100 44 Sweden

This paper introduces a "green" routing game between multiple logistic operators (players), each owning a mixed fleet of internal combustion engine vehicle (ICEV) and electric vehicle (EV) trucks. Each player faces the cost of delayed delivery (due to charging requirements of EVs) and a pollution cost levied on the ICEVs. This cost structure models: 1) limited battery capacity of EVs and their charging requirement;2) shared nature of charging facilities;3) pollution cost levied by regulatory agency on the use of ICEVs. We characterize Nash equilibria of this game and derive a condition for its uniqueness. We also use the gradient projection method to compute this equilibrium in a distributed manner. Our equilibrium analysis is useful to analyze the trade-off faced by players in incurring higher delay due to congestion at charging locations when the share of EVs increases versus a higher pollution cost when the share of ICEVs increases. A numerical example suggests that to increase marginal pollution cost can dramatically reduce inefficiency of equilibria. © 2022, CC BY-NC-ND.

关键词： Economic and social effects

Binary Bleed: Fast Distributed and Parallel Method for Automatic Model Selection

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Barron, Ryan Eren, Maksim E. Bhattarai, Manish Boureima, Ismael Matuszek, Cynthia Alexandrov, Boian S. Theoretical Division Los Alamos National Laboratory Los Alamos United States Advanced Research in Cyber Systems Los Alamos National Laboratory Los Alamos United States Department of Computer Science and Electrical Engineering University of Maryland Baltimore CountyMD United States

In several Machine Learning (ML) clustering and dimensionality reduction approaches, such as non-negative matrix factorization (NMF), RESCAL, and K-Means clustering, users must select a hyper-parameter k to define the number of clusters or components that yield an ideal separation of samples or clean clusters. This selection, while difficult, is crucial to avoid overfitting or underfitting the data. Several ML applications use scoring methods (e.g., Silhouette and Davies Boulding scores) to evaluate the cluster pattern stability for a specific k. The score is calculated for different trials over a range of k, and the ideal k is heuristically selected as the value before the model starts overfitting, indicated by a drop or increase in the score resembling an elbow curve plot. While the grid-search method can be used to accurately find a good k value, visiting a range of k can become time-consuming and computationally resource-intensive. In this paper, we introduce the Binary Bleed method based on binary search, which significantly reduces the k search space for these grid-search ML algorithms by truncating the target k values from the search space using a heuristic with thresholding over the scores. Binary Bleed is designed to work with single-node serial, single-node multi-processing, and distributed computing resources. In our experiments, we demonstrate the reduced search space gain over a naive sequential search of the ideal k and the accuracy of the Binary Bleed in identifying the correct k for NMFk, K-Means pyDNMFk, and pyDRESCALk with Silhouette and Davies Boulding scores. We make our implementation of Binary Bleed for the NMF algorithm available on GitHub. Copyright © 2024, The Authors. All rights reserved.

关键词： Non-negative matrix factorization