最新刊期
卷 11 ， 期 1 ， 2025
-
- 摘要：Data plays a crucial role in the development and application of artificial intelligence, which has become a consensus in industry and academia. Based on the interactive relationship between artificial intelligence and data, as well as the data-centric AI development practice, this paper proposes a data governance framework for artificial intelligence, which consists of six stages: source data governance, pre-trained data governance, evaluation data governance, fine-tuning data governance, inference data governance, and operation and maintenance data governance. Each stage has its key tasks and technologies. At the same time, this paper deeply analyzes the data governance cases and successful experiences of ChatGPT, Ziya2 and artificial intelligence models in the energy field to verify the effectiveness of the framework. The result shows that the framework plays an important role in improving the performance of artificial intelligence models and optimizing data governance processes. The framework provides reference for theoretical and practical innovation of data governance oriented to artificial intelligence.关键词：source data governance;pre-trained data governance;evaluation data governance;fine-tuning data governance;inference data governance;operation and maintenance data governance980|495|0更新时间：2025-01-15
- 摘要：Various departments of the enterprise implement decentralized management of data, and the chimney-style system construction causes data to be scattered in heterogeneous databases. Heterogeneous data poses a series of challenges to the current data integration work. In order to solve the problem of data aggregation and fusion of enterprise heterogeneous systems, an end-to-end data integration framework based on graph was proposed. Firstly, the table and field entity relationships were constructed into a network graph based on the primary and foreign key relationships of the relational data model. The table names and field names were regarded as different types of entities in the graph. Then, input the constructed graph into the graph neural network, and the vector representation of each node in the graph was obtained through graph convolution. Based on the node vectors, the node mapping relationship of any two graphs that need to be matched can be calculated. After aligning the tables and fields of the graph, the next step was to standardize the field values, meaning that the value of each cell was mapped to a standard value. Finally, engineer the above results into executable query statements for the database to achieve heterogeneous data fusion. Through verification on real data within the enterprise, the experimental results show that the framework proposed in the paper can improve the development efficiency of data integration, and the model is not limited by business fields and has strong portability.关键词：data integration;data fusion;heterogeneous data;schema matching;entity alignment;graph neural425|395|0更新时间：2025-01-15
- 摘要：With the popularization of generative artificial intelligence technology, the issue of copyright for AIGC is highlighted. The originality criterion under traditional “instrumental theory” and “contribution theory” is particularly inadequate when it is applied to AIGC, and it is necessary to construct more actionable originality criterion. By analyzing the legislative value of copyright law and empirical research on the norms of current laws, combined with empirical research on domestic and foreign judicial cases, the construction of originality criterion under “determinism” was proposed. Based on anthropocentrism, adhered to the uniqueness of the copyright subject, regarded labor input as the basis of rights, and the causal relationship theory in Anglo-American law system was introduced, which took the determination of the presentation results of human works and the control of the creative process as sufficient and necessary conditions. Thus, determinism is self-consistent in terms of objects, standards, laws, facts, and causal relationships, and can guide practical application.关键词：AIGC;originality criterion;determinism746|589|0更新时间：2025-01-15
-
- 摘要：In the digital age, business analytics courses require teaching reforms that align with contemporary developments, particularly the integration of emerging technologies (such as Python programming and big data analysis) into the curriculum. This paper, focusing on the course "Python programming for economics and finance", finds the probelm of existing courses, including overly theoretical content, the singleness of assessment methods and inadequate cultivation of students' comprehensive application abilities. A series of reform measures are implemented, including curriculum content updates and innovative teaching methods. Comparisons between pre-and-post reform questionnaires and assessments reveal that these measures effectively enhance students' programming and problem-solving abilities, indicating the success of this educational reform and providing valuable insights for related research.关键词：teaching reform;big data;economic management563|993|0更新时间：2025-01-15
- 摘要：With the rapid advancement of big data industry, professional education pertaining to big data has garnered substantial global recognition. However, the gap between practical teaching methodologies and real-world application scenarios represents a significant hurdle in fostering skilled big data professionals. To address this challenge, this paper proposes the construction of a comprehensive digital case library tailored for core courses in big data. This library, alongside innovative integration strategies, aims to bridge the gap between theory and practice. This paper outlines a structured approach to establish a systematic and serialized data resource case library, adopting a dynamic hierarchical design that comprehensively illustrates the entire big data processing workflow while enabling dynamic expansion. The objective is to contribute to the enhancement of big data curricula and practical teaching, ultimately empowering the practical abilities cultivation of big data professionals. Taking the digital case study of "high-risk population cardiovascular and cerebrovascular disease" as an example, this paper delves into the curriculum model and implementation strategies that seamlessly integrate case resources with theoretical teachings. Through collaboration with a national data center, we achieve a close alignment between practical teaching and real-world application scenarios. The construction of this caselibrary resource will bolster the theoretical acumen and practical proficiencies of big data professionals.关键词：digital case;big data professional;practical teaching;dynamic hierarchical design339|779|0更新时间：2025-01-15
- 摘要：In response to the teaching challenges in big data practice education, such as building big data cluster environments, environment management, component compatibility, and hardware maintenance, Huawei MRS big data platform is used for course exploration as a practical teaching platform. The advantages of the Huawei MRS big data platform in teaching were discussed. The reform concept of “learning, teaching, practicing and thinking” was proposed, which included theoretical teaching, experimental teaching, comprehensive practice and curriculum evaluation. The paper discussed the implementation process of curriculum reform from the aspects of knowledge system, experimental teaching, comprehensive practice tasks and course assessment methods. This course has been demonstrated at Huawei "Intelligent Foundation" industrial-education integration collaborative education base, effectively improving students' independent research capabilities and innovation skills of big data technology.关键词：Huawei MRS platform;big data teaching;curriculum reform250|376|0更新时间：2025-01-15
COLUMN
- 摘要：Large language models have shown exceptional performance in complex tasks, such as text and visual processing, drawing significant attention from industry and academia. Both the training and inference phases of large language models are heavily reliant on GPU computational power. However, the limited capacity and volatility of GPU memory pose significant challenges in meeting the storage demands during these phases. This paper conducts an in-depth analysis of the challenges faced by storage systems in the era of large language models: (1) the data of large language models exhibits a high degree of fragmentation, and sparse semantic representation further reduces storage system efficiency; (2) large language model training and inference demand high data read/write bandwidth, but the high communication overhead associated with data transfer across heterogeneous storage media complicates the expansion of GPU memory using these media; (3) the fault tolerance requirements during large language model training are stringent, yet directly applying CPU-centric fault tolerance techniques incurs prohibitive costs. To address these challenges, existing solutions are summarized from three perspectives: data management, storage expansion, and fault tolerance. Finally, the future development trends of storage systems in the era of large language models are discussed.关键词：large language model;storage system;data management;storage scalability;data fault tolerance420|441|0更新时间：2025-01-15
- 摘要：Data interoperating over Internet enables cross-domain connectivity and exchange of data, where authorization is a basic mechanism to protect data ownership and ensure data security. The complexity of cross-domain data exchange poses higher requirements for authorization technology in terms of utility and credibility, thus, there is an urgent need to carry out research on authorization technology for data interoperating-oriented. Firstly, the basic concepts and knowledge of authorization technology were briefly introduced, and the technical requirements of data interoperating-oriented for authorization were analyzed. Secondly, the existing authorization technical capabilities were analyzed from there aspects: trust mechanism, permission model, and policy management. Finally, in response to the limitations of existing authorization technology capabilities, some potential research questions were proposed, aiming to provide ideas and references for follow-up research.关键词：data interoperating;cross-domain data exchange;authorization protocol;trust mechanism220|393|0更新时间：2025-01-15
- 摘要：The visual enhancement technology of sandstorm images aims to improve the visual perception clarity of the captured data by imaging equipment under sandstorm weather, assisting high-level vision algorithms to enhance the ability to obtain key features from the data. With the increasing frequency of sandstorm weather, sandstorm image visual enhancement technology has gradually become a research hotspot in the field of image processing and has a very wide range of applications in AI-related industries， such as remote sensing surveys, unmanned driving, and smart transportation. To better understand the current research progress in the field of sandstorm image processing, this paper elaborates in detail on sandstorm image enhancement algorithms based on traditional technology, sandstorm image restoration algorithms based on atmospheric scattering models, and summarizes data-driven sandstorm image reconstruction algorithms. It also conducts a detailed study and analysis of the construction of current public datasets and briefly sorts out the evaluation indicators commonly used in the image enhancement field. Finally, a summary of the above content is made, pointing out the difficulties existing in current research and prospects for the development direction of future sandstorm image enhancement technology.关键词：image enhancement;sand-dust image;computer vision;deep learning304|309|0更新时间：2025-01-15
- 摘要：In practical applications, purely periodic data is relatively rare, and most data often exhibit non-periodic characteristics, making it challenging to predict or describe through simple periodic changes. Single neural network frequently encounters issues such as overfitting, difficulties in capturing long-term dependencies, and limited ability to capture nonlinear relationships when dealing with non-periodic time series. To effectively forecast non-periodic time series, the ILTNet model was proposed based on the Informer model. The ILTNet model integrated linear prediction (AR model) and nonlinear prediction (Informer model and recurrent skip components), enabling effective capture of long-term dependencies. Experiments show that ILTNet model demonstrates significant advantages in non-periodic time series forecasting compared to the LSTNet, Informer, AR, and GRU models. For example, on the Exchange Rate dataset, ILTNet model reduces RSE by 0.0333 and 0.0277 compared to LSTNet model at horizons of 96 and 128, respectively.Compared to Informer model, ILTNet model achieves significant RSE reductions at all horizons, particularly reducing RSE by 0.2877 at a horizon of 96.关键词：time series forecasting;convolutional neural network;LSTNet;Informer314|573|0更新时间：2025-01-15
-
Query scheduling based on cloud-edge multi-data warehouse architecture and cost prediction model AI导读
摘要：With the development of cloud computing and big data, traditional local data warehouses are difficult to expand and have low data processing efficiency, As a result, the data warehouse of cloud edge architecture comes into being. The architectures data warehouses are distributed in the cloud center and at the edge, making data storage and processing more flexible and supporting business such as data security, data privacy and cross-geographic data sharing while ensuring query efficiency. This paper designed a scheduling framework based on cloud edge multi-data warehouses, integrated the query cost prediction model with machine learning technology as the core, and realized cloud edge collaborative execution and cloud edge selective execution on multiple query granularity, so as to improve the performance and query efficiency of the whole system. In addition, the multi-feature fusion and feature selection method is proposed to enhance the query cost information. The scheduling framework and optimization algorithm achieve significant performance improvement on SSB and TPC-DS datasets. It provides an effective solution for the query scheduling of data warehouse under the cloud edge multi-data warehouses architecture.关键词：scheduling framework;query cost prediction;random forest;feature selection171|263|0更新时间：2025-01-15
STUDY
-
Analysis and exploration of algorithmic ethical difficulties in generative artificial intelligence AI导读
摘要：Since 2022, generative artificial intelligence technology and industry have been developing rapidly. This paper focused on the generative algorithms used in generative AI technology, and proposed the "Three Laws of Generative Algorithms" ethical principles in combination with relevant regulations. Combined with its technical characteristics, it analyzed the ethical difficulties that exist in practice, and then explored some preliminary solution frameworks.关键词：large language model;generative algorithm;algorithmic ethics657|754|0更新时间：2025-01-15
FORUM
- 摘要：Knowledge graphs organize and represent entity relationships through graph structures, providing a foundation for machine understanding and reasoning, but their reasoning capabilities are limited by coverage and manual rules. Large language models demonstrate strong semantic understanding and generation abilities but lack effective utilization of symbolic knowledge and interpretability. To combine the strengths of both technologies, academic and industrial communities have devoted significant effort in recent years to exploring the integration of knowledge graphs and large language models, aiming to build more powerful and interpretable AI systems. Firstly, this paper reviews the current state of research on the fusion of knowledge graphs and large language models, with a focus on the key achievements in enhancing fact representation and logical reasoning. These achievements include pre-trained language models based on knowledge graphs, knowledge graph representation learning based on large language models, and reasoning methodsthat leverage the fusion of the two approaches. Furthermore, the paper outlines the mainstream technical approaches and application scenarios of graph-model integration in the industry. Finally, future development directions of graph-model intgeration are discussed, and it is posited that the integration of these two technologies represents a crucial trend in the advancement of artificial intelligence.关键词：large language model;graph neural network;machine reasoning;pre-training;attention;graph representation learning408|385|0更新时间：2025-01-15
-
- 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