Please wait a minute...
Most Download
Most Cited
Links
Visited
Total visitors:
Visitors of today:
Now online:

Current Issue

    15 September 2023, Volume 5 Issue 3
    Review Intelligence
    Parallel Sharhili: a new approach to sustainable development and intelligent management of ecological resources
    Peiding PI, Qinghua NI, Jing YANG, Mengzhen KANG, Xuanhao LI, Yingkun DU, Fei-Yue WANG
    2023, 5(3):  283-292.  doi:10.11959/j.issn.2096-6652.202325
    Asbtract ( 293 )   HTML ( 52)   PDF (40201KB) ( 394 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Sharhili Nature Reserve is a natural treasure which enjoys both rare biological resources and incalculable ecological value, and it is known as “China’s last pure land” and “rare natural gene pool”.To ensure the sustainable development of this natural treasure, this study proposes the concept of “Parallel Sharhili”, which comprehensively assists in the precise management of ecological resources through intelligent means.The core idea involves utilizing “artificial systems” for description to build high-precision models, “computational experiments” for prediction to analyze ecosystem changes, and “parallel execution” for prescription or control and management to scientifically formulate conservation strategies.This study explores an intelligent management model for nature reserves, aiming at promoting the sustainable development of the reserve and its precious biological resources.

    Surveys and Prospectives
    A survey of deep learning-based MRI stroke lesion segmentation methods
    Weiyi YU, Tao CHEN, Junping ZHANG, Hongming SHAN
    2023, 5(3):  293-312.  doi:10.11959/j.issn.2096-6652.202328
    Asbtract ( 827 )   HTML ( 174)   PDF (23370KB) ( 597 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Automatic stroke lesion segmentation has become a research hotspot in recent years.In order to comprehensively review current progress of deep learning-based MRI stroke lesion segmentation methods, start with the clinical problems of stroke treatment, we further elaborate the research background and challenges of deep learning-based lesion segmentation, and introduce common public datasets (ISLES and ATLAS) for stroke lesion segmentation.Then, we focus on the innovation and progress of deep learning-based stroke lesion segmentation methods, and summarize the research progress from three perspectives: network structure, training strategy, and loss function, and compare the advantages and disadvantages of various methods.Finally, we discusse the difficulties and challenges in this research and its future development trend.

    Survey on multi-agent reinforcement learning methods from the perspective of population
    Fengtao XIANG, Junren LUO, Xueqiang GU, Jiongming SU, Wanpeng ZHANG
    2023, 5(3):  313-329.  doi:10.11959/j.issn.2096-6652.202326
    Asbtract ( 916 )   HTML ( 98)   PDF (3737KB) ( 1385 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Multi-agent systems are a cutting-edge research concept in the field of distributed artificial intelligence. Traditional multi-agent reinforcement learning methods mainly focus on topics such as group behavior emergence, multi-agent cooperation and coordination, communication and communication between agents, opponent modeling and prediction. However, they still face challenges such as observable environment, non-stationary opponent strategies, high dimensionality of decision space, and difficulty in understanding credit allocation. How to design multi-agent reinforcement learning methods that meet the large number and scale of intelligent agents and adapt to multiple different application scenarios is a cutting-edge topic in this field. This article first outlined the relevant research progress of multi-agent reinforcement learning. Secondly, a comprehensive overview and induction of multi-agent learning methods with multiple types and paradigms were conducted from the perspectives of scalability and population adaptation. Four major categories of scalable learning methods were systematically sorted out, including set permutation invariance, attention, graph and network theory, and mean field theory. There were four major categories of population adaptive reinforcement learning methods: transfer learning, course learning, meta learning, and meta game, and typical application scenarios were provided. Finally, the frontier research directions were prospected from five aspects: benchmark platform development, two-layer optimization architecture, adversarial strategy learning, human-machine collaborative value alignment and adaptive game decision-making loop, providing reference for the research on relevant frontier key issues of multi-agent reinforcement learning in multimodal environments.

    Special Column: Intelligent Technology and Social Computing
    Analysis and prediction of GitHub company influence based on machine learning
    Mingyu WANG, Qingyuan GONG, Jingjing QU, Xin WANG
    2023, 5(3):  330-342.  doi:10.11959/j.issn.2096-6652.202327
    Asbtract ( 311 )   HTML ( 45)   PDF (19716KB) ( 516 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    The influence of a company is not only related to its industry competitiveness, but also affects its public reputation and future development.However, there has been no unified standard for evaluating the influence of a company.GitHub is a representative open-source platform for software development code repositories.Existing research typically used the total number of stars a company receives for projects posted on GitHub to measure its influence, but this approach is difficult to measure the potential of small, micro, and nascent companies.The paper predicted the future influence level of a company by introducing the scientist's influence measure h-index, using GitHub as the information source, and modeling the company network.Features was extracted features based on this network to build the classifier, which predicted the future influence level of the company.The SHAP model explanation technique was further applied on this basis to identify the important features that determined the influence of a company.The experimental results showed that the XGBoost model achieved an accuracy of 0.92 and an average AUC of 0.93 on the real-world GitHub dataset.In summary, the proposed method could accurately and reliably predict the influence of companies.

    Research on the overseas communication effect of Chinese realistic theme TV series: take YouTube platform as an example
    Xue MENG, Ruonan YANG, Ruiqi LI
    2023, 5(3):  343-351.  doi:10.11959/j.issn.2096-6652.202329
    Asbtract ( 353 )   HTML ( 43)   PDF (5290KB) ( 701 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    YouTube is the international video website with the largest number of active users in the world, and it is also one of the main channels for Chinese film and television content to go global.Natural language processing technology was used to analyze the communication effect of Chinese realistic theme TV series on the YouTube platform.The factors affecting the communication effect of Chinese realistic theme TV series on YouTube could be found from different dimensions such as content, transmitter, and recipient.Then more works can be guided to accurately disseminate.It was found that the acceptance of overseas audiences to the work was affected by the type, content, team and translation of realistic TV series.The communication efficiency of Chinese realistic theme TV series on overseas platforms could be improved only by grasping user preferences, strengthening brand awareness, understanding channel characteristics and attaching importance to subtitle translation.The accurate and efficient dissemination of “Chinese-style modernization” stories on international video platforms could be realized.

    An empirical study on the impact of enterprise digital transformation on employment scale and structure
    Jiting HUANG, Kexin GUO, Jiayin QI
    2023, 5(3):  352-365.  doi:10.11959/j.issn.2096-6652.202336
    Asbtract ( 411 )   HTML ( 59)   PDF (4075KB) ( 469 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    With the advent of the digital age, people are paying more and more attention to the employment scale and employment structure of enterprises.Used the enterprise panel data of listed companies from 2011 to 2019, the impact of enterprise digital transformation on employment scale and employment structure was studied deeply by the fixed-effect model and mediation effect model.The results show that, firstly, the digital transformation of enterprises significantly expand the employment scale of enterprises.At the same time, the impact of digitalization on the job structure has a strong"knowledge bias", showing a bias towards high skills and knowledge, and an increase in the demand for talents who master knowledge related to cutting-edge industries.The results of the intermediary effect test show that the digital transformation of enterprises indirectly promotes the employment scale of enterprises through market scale effect and product innovation.The results of heterogeneity analysis show that the high degree of digitalization of the environment and the degree of digitalization of non-state-owned enterprises have a more significant impact on the employment structure.

    Game patrol strategy for hazardous gas leakage in chemical parks
    Yin CHEN, Lize ZHANG, Guohua SHUAI, Lili CHEN, Zhen WANG
    2023, 5(3):  366-377.  doi:10.11959/j.issn.2096-6652.202338
    Asbtract ( 176 )   HTML ( 15)   PDF (33637KB) ( 289 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    In the context of promoting the integration of smart cities, the pace of construction of smart chemical parks is gradually accelerating.Since accidents in China’s chemical industry have occurred frequently in recent years, causing great damage and loss to public safety, improving the safety management and emergency response capability of chemical parks is an urgent need to be solved.For this kind of safety problems, this paper proposed a game patrol strategy for hazardous gas leakage in chemical parks.Firstly, the convective diffusion model was used to describe the process of hazardous gas leakage.Secondly, game theory was introduced to model the confrontation process between the attacking and defending parties in the patrol problem, and the response time of the defenders to a safety incident was correlated with its benefit.Then, a multilinear programming-based GGC algorithm was proposed to solve this game model.Finally, the gains of the game model in this paper were compared with the other two basic methods in the scenarios of three real chemical park case with different sizes.The results show that the model can effectively improve the gains of the defender and reduce the gains of the attacker.

    Special Topic: Diffusion Model and Artificial Intelligence Generated Content
    Data augmentation method based on diffusion model for domain generalization
    Yujun TONG, Heqing WANG, Yueheng LUO, Wenxin NING, Mandan GUAN, Wenqing YU, Keyan HUANG, Jiaxun ZHANG, Zhanyu MA
    2023, 5(3):  380-388.  doi:10.11959/j.issn.2096-6652.202334
    Asbtract ( 655 )   HTML ( 65)   PDF (7426KB) ( 402 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Domain generalization is an important and challenging problem in computer vision, arising from the distribution shift of real-world data.In practical applications, it is common to encounter training and testing data from different domains, and the difference in data distribution can lead to performance degradation during testing.In this paper, we propose a domain generalization method based on latent space data augmentation.Unlike traditional image-level data augmentation approaches, the method introduces a diffusion model in the latent space to achieve fine control and diversity generation of features, thereby achieving feature level data augmentation and enhancing the model's generalization ability in the target domain.Specifically, the classifier-based implicit diffusion model, trained within the latent space, can conditionally generate accurate and rich source domain features.It leverages efficient sampling techniques to expedite the generation of augmented features.Experimental results show that the method has achieved significant performance improvement in various domain generalization tasks, and has good effectiveness and robustness in real scenarios.The key innovation of this paper lies in shifting data augmentation to the latent space level and introducing the diffusion model for augmentation, providing a novel approach to address the domain generalization problem.

    Fine-grained urban flow inference based on diffusion models with incomplete data
    Yuhao ZHENG, Senzhang WANG
    2023, 5(3):  389-396.  doi:10.11959/j.issn.2096-6652.202330
    Asbtract ( 277 )   HTML ( 19)   PDF (9260KB) ( 521 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    To obtain detailed traffic flow data for each road segment of the city, it is necessary to deploy a large number of sensing devices and dense observation stations, which increases the costs of daily operations and equipment maintenance.At the same time, traditional traffic flow survey techniques are noisy and inaccurate, and the reliability of the detected data results is not guaranteed.Therefore, inferring fine-grained urban traffic flow based on coarse-grained and noiseinclusive sensor observations has become an important research topic.To address the above problems, we proposed a denoising diffusion model based on spatio-temporal attention, with the intention of providing fine-grained urban traffic base data in different scenarios of traffic demand, and laying the foundation for traffic planning and intelligent transportation system construction.

    Lipsynthesis incorporating audio-visual synchronisation
    Cong JIN, Jie WANG, Zichun GUO, Jing WANG
    2023, 5(3):  397-405.  doi:10.11959/j.issn.2096-6652.202335
    Asbtract ( 446 )   HTML ( 59)   PDF (13372KB) ( 429 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    With the flourishing development of video-based information dissemination, audio and video synchronization is gradually becoming an important standard for measuring video quality.Deep synthesis technology has been entering the public's view in the international communication field, and lip-sync technology integrating audio and video synchronization has attracted more and more attention.The existing lip-synthesis models are mainly based on lip-synthesis of static images, which are not effective for synthesis of dynamic videos, and most of them use English datasets for training which results in poor synthesis of Chinese Mandarin.To address these problems, this paper conducted optimization experiments on the Wav2Lip lip synthesis model in Chinese context based on its research foundation, and tested the effect of different routes of training models through multiple sets of experiments, which provided important reference values for the subsequent Wav2Lip series research.This study realized lip synthesis from speech-driven to text-driven, discussed the application of lip synthesis in multiple fields such as virtual digital human, and laid the foundation for the broader application and development of lip synthesis technology.

    AI-driven digital image art creation: methods and case analysis
    Changsheng WANG
    2023, 5(3):  406-414.  doi:10.11959/j.issn.2096-6652.202333
    Asbtract ( 1218 )   HTML ( 308)   PDF (48558KB) ( 853 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    In recent years, the rapid advancement of artificial intelligence technology has introduced novel tools for the realm of digital image creation.The use of AI tools for digital image creation has gradually become a common practice.In this context, this paper analyzed the methods and cases of AI-driven digital image art creation through interviews, surveys, and case analyses.The survey revealed that the majority of respondents believed that AI tools had a significant impact on their creative process, especially the widespread use of direct output from artificial intelligence methods in digital image art creation.By analyzing the cases in detail, this paper revealed the typical paradigms of AI-based digital image art creation and summarized them as the “AI art creation workflow”, which included AI direct output, AI assisted drawing, and ControlNet precision drawing.These methods can effectively replace repetitive steps in digital image creation and improve image output efficiency, assist in digital image drawing and break through the conventional drawing ideas, and achieve precise and controllable expression of digital image art content.

    Potential risks and governance strategies of artificial intelligence generated content technology
    Yaling LI, Yuanqi QIN, Que WEI
    2023, 5(3):  415-423.  doi:10.11959/j.issn.2096-6652.202332
    Asbtract ( 753 )   HTML ( 95)   PDF (2926KB) ( 816 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Artificial intelligence generated content technology is reshaping the production and consumption patterns of digital content.The research was based on the basic concepts and technological development process of artificial intelligence generated content technology, focusing on exploring ethical risks and governance challenges such as privacy leakage, algorithm hegemony, digital divide, intellectual property protection and employment impact caused by artificial intelligence generated content.After reviewing the typical governance paths and experiences of developed countries abroad, this paper summarized and proposed measures and development suggestions for China to address the potential risks of AI generated content.One was to prioritize development and coordinate layout; Second was to strengthen the full factor investment in the research and development of artificial intelligence technology; Third was to improve the governance rule system of artificial intelligence technology; Fourth was to accelerate the research and development of artificial intelligence governance technology.

    Future of AI painting from a legal perspective: the balance between freedom and control
    Yinyuan MA
    2023, 5(3):  424-430.  doi:10.11959/j.issn.2096-6652.202331
    Asbtract ( 491 )   HTML ( 267)   PDF (1501KB) ( 333 )   Knowledge map   
    References | Related Articles | Metrics

    AI painting, as a representative of the new Internet technology, has a free environment consistent with the development of the early Internet.However, the free sharing of public resources is not eternal.The conflict between old system and new will inevitably brings many problems in the application of the existing laws.China’s existing laws and regulations for network service providers responsibility than the outside to be more perfect, again forced to increase the service provider or to serve the interests of the users responsibility will lead to imbalance, and there is no necessity for reference.Under the existing legal framework, the use of special legislation and industry autonomy to reinforce the power limits of the dominant player is enough, and the future of AI painting needs to always pay attention to the value of balance.