最新刊期
卷 11 ， 期 3 ， 2025
-
- 摘要：With the rapid development of new generation information and communication technologies such as artificial intelligence, the wide-area distributed intelligent computing environment has become an important infrastructure. In response to the challenge of efficient collaborative scheduling of resources in a wide-area distributed intelligent computing environment, a runtime collaborative scheduling of resources in computing power network method is proposed. The key task selection and backfilling, network traffic scheduling of core flow, and adaptive data placement strategies are designed. The usage of computing, network, and storage resources in the computing power network are comprehensively analyzed, and the strategies are collaboratively implemented to optimized the global utilization of resources. The experimental results indicate that the proposed method can effectively improve the system throughput while optimizing the global data migration cost compared with the existing methods.关键词：task scheduling;network resource scheduling;data placement;collaborative scheduling490|313|0更新时间：2025-05-15
-
SpanTrain: a cross-domain distributed model training system for cloud-edge-end heterogeneous devices AI导读
摘要：Currently, in addition to cloud computing centers, the edge and end environments represented by the internet of things, fixed or mobile computing edges are also filled with a large number of intelligent computing devices. Expanding the deep neural network (DNN) training from cloud computing centers to the edge and end has significant advantages in aspects such as support for new application patterns, protection of data privacy, and control of training costs. Most existing distributed training systems are designed for homogeneous devices, and they are difficult to adapt to the heterogeneous computing environments of cloud-edge-end. For this reason, a cross-domain distributed training system named SpanTrain, which is based on the heterogeneous devices of cloud, edge, and end, has been designed. Through a novel hybrid pipeline parallel mechanism, it realizes the efficient DNN model training with the collaboration of the heterogeneous devices of cloud, edge, and end. Moreover, experiments have been carried out in an environment containing typical heterogeneous devices. Experiments in typical cloud-edge-end heterogeneous environments demonstrate that SpanTrain achieves 1.17x~3.15x higher training throughput compared to state-of-the-art systems, while improving resource utilization of heterogeneous devices by 39%. These results validate the efficiency of SpanTrain for DNN training in cloud-edge-end heterogeneous environments.关键词：distributed computing;DNN training;parallel training strategy384|322|0更新时间：2025-05-15 - 摘要：The cloud-edge-device collaborative architecture serves as the foundation for applications such as smart manufacturing and smart cities, with collaborative computing as a critical supporting technology. Within this architecture, due to the diversity and large scale of data, traditional data management techniques struggle to support real-time querying of large-scale data. To address this, this paper proposes a database pre-aggregation method tailored to cloud-edge-device collaboration, utilizing an automated materialized view generation strategy to achieve data pre-aggregation and improve real-time query performance. Experimental results show that this pre-aggregation method reduces query time by up to 68.45%, significantly enhancing data query performance in cloud-edge-device collaborative architecture.关键词：cloud-edge-device collaborative;materialized view;deep reinforcement learning152|149|0更新时间：2025-05-15
- 摘要：The LSM-tree based key-value store has become an ideal choice for cloud and edge data management due to its efficient data storage mechanism. However, the leveled compaction strategy used in LSM-tree suffered from high write amplification, which negatively impacted foreground write performance. Reducing write amplification and further improving write performance without sacrificing read performance become a major challenge in optimizing LSM-tree. To address this issue, we proposed a novel key-value store system, LooseKV, which leveraged the tiered compaction strategy to significantly reduce write amplification. Additionally, LooseKV introduced a skip-list-based memory index, combined with a lightweight index update strategy and integration with iterators in the tiered structure, effectively improving the read performance issues associated with the tiered strategy. Experimental results demonstrated that LooseKV achieved 1.18 to 2.28 times higher random write throughput than LevelDB, 1.01 to 1.26 times better random read performance, and slightly lower sequential read performance compared to LevelDB, but comparable to PebblesDB.关键词：log-structured merge tree;write amplification;read performance optimization152|153|0更新时间：2025-05-15
- 摘要：In video analysis, data drift is one of the main issues affecting system inference performance. To address this, this paper proposes a mobile-oriented edge-cloud collaborative continuous learning framework. Specifically, this framework deploys a model caching on resource-constrained mobile devices and designs an edge-cloud collaborative hierarchical adaptation strategy that integrates “reuse, fine-tuning, and retraining”. Constructing a prefix tree storage structure based on domain attributes enables rapid model retrieval, thereby reducing cloud queuing delays during adaptations. Evaluation results demonstrate that this framework outperforms existing state-of-the-art solutions in object detection and image classification tasks, achieving a 34.5-times reduction in reuse response time and a 1.99-times reduction in retraining time while achieving optimal overall accuracy in multi-device system adaptations. This highlights the framework’s significant potential for enhancing response speed and system scalability.关键词：device-cloud collaboration;video analytics;edge computing;continuous learning151|146|0更新时间：2025-05-15
TOPIC
- 摘要：The public data classification and grading system is an important part of the data foundation system. Its strong and orderly application and implementation of data elements of the system supply, efficient circulation and security governance has an important role in supporting and guaranteeing. This paper proposes a feasible closed-loop management method for public data classification and grading based on the pilot situation of classification and grading of relevant departments and cities in Jiangsu Province. We establish a classification and grading control system for data of different sensitive levels. Under the premise of safeguarding data security, we optimize the allocation of data resources, promote data sharing and opening and authorized operation, and improve the security and transparency of data circulation and transaction, so as to promote the effective release of the value of data elements.关键词：public data;data element;security governance;classification and grading closed-loop management;classification and grading control system396|371|0更新时间：2025-05-15
- 摘要：Various regions and departments are actively exploring new models of digital government construction, promoting the expansion of electronic license applications and national mutual recognition. Shandong Province focuses on deepening the reform of “streamlining administration, delegating powers, and improving services” to optimize the business environment, and regards the construction of an “unproven province” as an important measure to promote the government's digital transformation. However, Shandong is in the early stage of building an “unproven province”, and there are still some problems and challenges in exploring “unproven” services. In response to this, the difficulties, pain points, and bottlenecks were analyzed, and targeted countermeasures and suggestions were put forward, aiming to accelerate the process of building a digital government in Shandong and further enhance the efficiency of government services.关键词：electronic certificate proof;digital government209|191|0更新时间：2025-05-15
- 摘要：The orderly advancement of data circulation and the effective regulation of the data market through coordinated deployment are the important issue. Addressing challenges such as the low identifiability of data circulation targets and the difficulties in managing the negative externalities of data, this paper defined the forms of data circulation targets from the perspectives of measurement, transmission, and security, proposing identifiable boxed data products as circulation targets. Furthermore, based on these boxed data products, the contents involved in the regulatory system for data product circulation were clarified and a practical regulatory framework for data products was introduced. This framework integrates regulatory indicators at each stage of design, production, and circulation, enabling rapid and effective identification and management of circulation risks.关键词：data product;circulation regulatory;data market248|216|0更新时间：2025-05-15
FORUM
- 摘要：Embodied agents refer to intelligent entities capable of completing one or multiple tasks based on instructions and possessing the ability to interact with the physical environment. These agents have immense potential applications across various fields, such as service robotics, intelligent education, and assistive healthcare, and represent a crucial pathway toward realizing general-purpose robots. With the advancement of multimodal large language models, embodied agents possess enhanced abilities in natural language understanding, reasoning, and environmental perception, significantly accelerating progress in this domain. Although many outstanding works have emerged in recent years, the field still lacks comprehensive surveys and targeted evaluations. To help researchers quickly and thoroughly know the developments in this area, in-depth review and analysis were conducted. Multimodal large language models were introducted, followed by datasets and a review of the physical carriers used for constructing embodied intelligent agents. Then, three key research directions are analyzed, including embodied large models, high-level task planning, and low-level action control. Finally, the challenges and limitations of embodied agents were summarized and potential future directions were explored. This review serves as a foundational reference for the research community and fosters further development and innovation in the field.关键词：embodied agent;multimodal large language model;robot;vision-language model;embodied intelligence1133|521|0更新时间：2025-05-15
- 摘要：Massive data and artificial intelligence technology provide great convenience for communication research, which can effectively solve the problem of communication effect measurement. Existing research in this field mainly focuses on the research paradigm, scenario application and development path of computational communication, and lacks the sorting and categorization of novel computational method in computational communication. This paper adopted the literature research method, proposed a framework for categorizing computational method based on the 5W communication model, and systematically compiled the new computational method in the communication process with the goal of improving the communication effect, describing the ideas of the different computational method, the required data sets and the evaluation methods. This paper analyzed the challenges faced by computational communication technology and looked forward to possible future research directions.关键词：computational communication;5W communication model;computational communication technology;computational method;social network395|1486|0更新时间：2025-05-15
- 摘要：Deepfake technology poses a serious threat to social economy, political stability, and social security. Among them, voice forgery technology is widely used in harmful activities such as phone scams and public opinion manipulation. In recent years, with the application of deep learning technology, speech synthesis and voice conversion technology have advanced rapidly, and can generate fake voices which are enough to deceive both machines and humans. In response to the harm of voice forgery, many voice deception detection technologies have emerged to improve the reliability of speaker verification systems. However, existing methods often rely on prior knowledge of known attack types, and their generalization ability is limited when they faced with unknown attack types of prior knowledge. We built a voice deception detection system based on One-Class learning, which enhanced the generalization ability of the model by establishing strict decision boundaries for real voices and judging samples outside the boundaries as fake voices. In addition, to address the scarcity of fake voice data, the more versatile and robust self-supervised model Wav2vec2 was introduced for feature extraction, further improving the recognition accuracy of the model when faced with unknown attacks. Experimental results show that the proposed method can not only ensure good voice anti-spoofing performance, but also reduce the potential interference of the CM system to the downstream ASV system, effectively solving the problems of scarce fake voice data and insufficient generalization ability of the model.关键词：spoof detection;One-Class learning;self-supervised learning111|198|0更新时间：2025-05-15
- 摘要：Under the impetus of multi-bilateral economic dispatch, the peer-to-peer energy trading mode has gained significant attention. However, data privacy breaches and transaction regulatory challenges have emerged as critical barriers to its growth. A new peer-to-peer energy transaction matching mechanism named PV-RCI was proposed, which ensured the effectiveness of privacy-preserving mechanism and the validity of transaction matching mechanism. By incorporating a trusted auditing party and leveraging homomorphic encryption in conjunction with zero-knowledge proof, the proposed mechanism not only safeguarded data privacy but also ensured the verifiability of transaction matching outcomes. On the basis of promoting transaction matching between supply and demand parties, the proposed mechanism significantly reduced the overall social cost of transaction, offering a new approach to addressing privacy- preserving and regulatory issues in the energy trading sector.关键词：privacy-preserving;transaction matching;zero-knowledge proof85|149|0更新时间：2025-05-15
STUDY
- Postal code：100079
- Tel：（010）53879208、53859533 Email：bdr@bjxintong.com.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
- Total Visits：273110 Visits Today：16
0