On March 27, the Cloud Forum on "Granular Computing and Uncertainty Information Processing" was successfully held by CAAI and its Special Committee on Granular Computing and Knowledge Discovery and the School of Computer and Cyberspace Security of Hebei Normal University. Guests from academia and industry expressed their views and launched a heated discussion on the cutting-edge theories, technologies and applications of uncertain information processing in the field of artificial intelligence from the perspective of granular computing, which attracted 293,025 online views and became a remarkable cloud forum held by the CAAI in 2021.
The host of this forum is Professor Zhengsheng MI from Hebei Normal University and the deputy director of CAAI Special Committee on Granule Computing and Knowledge Discovery. First of all, Professor Duoqian MIAO from Tongji University, director of the CAAI Special Committee on Granule Computing and Knowledge Discovery, delivered a speech. Subsequently, seven experts gave keynote speeches. The first half of the conference was hosted by Professor Duoqian MIAO of Tongji University, the chairman of the forum, and the second half was hosted by Professor Shuliang ZHAO of Hebei Normal University, the chairman of the forum.
During the presentation session, Licheng JIAO shared a report entitled "Image Interpretation Based on Brain-like Perception and Cognition" , he expounded the theory and method of acquiring, modeling and applying high-resolution remote sensing images from three aspects of human cognitive features, namely "sense, knowledge and application" , and shared the team's research results, finally, he shared his views on the six major fields of cognitive science and the challenges they faced.
Professor Zhiguo GONG, director of the Department of Computer and Information Science at the University of Macau, proposed the spatio-temporal structure of social media with the theme of "Mining and Analysis of Social Media Data". He shared the boosting effect of social media data mining and analysis technology on Macao's tourism development from aspects of spatiotemporal data acquisition, hot spot identification, online event discovery and understanding, and cross-city scenic spot recommendation.
Qinghua HU said in the keynote report of "Multi-granularity Learning of Long-Tail Distribution Data" that the classification tasks faced by artificial intelligence are becoming increasingly large, and the processing of long-tail distribution data is crucial in the application fields such as unmanned driving and intelligent diagnosis. He proposed four solutions from the perspective of long-tail multi-granularity modeling: feature grouping modeling considering hierarchy, multi-granularity knowledge transfers from coarse to fine, multi-granularity classification modeling based on hierarchy, and tail-category modeling guided by hierarchy.
Focusing on the theme of "the theory of learnability from the perspective of stochastic consistency”, Qian pointed out that in the process of human learning, as a result of the learner's knowledge defect or the noise interference in the data, there may be random consistency in the learning results, which leads to the lack of objectivity, interpretability and repeatability in the decision making; it brings challenges to the classical learnable theory and method. In order to alleviate the random consistency in learning process, he put forward the theory and methods of probability approximate correctness, optimization theory, assumption space responsibility theory and test evaluation.
Yue Xiaodong, member of the CAAI Machine Learning Committee, standing member of the CAAI Granular Computing and Knowledge Discovery Committee, associate professor of Shanghai University, a deputy director of the Department of Intelligence Science of the School of Computer Science, and deputy director of the Artificial Intelligence Committee of the Shanghai Computer Society, focused on the theme of " Field-oriented machine learning and its application in medical assisted diagnosis ". Starting from the practical problems of machine learning in the field of medical application, he discussed the opportunities and challenges of field-oriented machine learning combined with clinical cases of tumor disease auxiliary diagnosis. He introduced and analyzed the machine learning theory and engineering practice about integrating domain elements, needs, and constraints from the perspectives of feature representation, classification strategy, model interpretation and domain knowledge fusion.
Li Jinhai, a standing committee member of CAAI Granular Computing and Knowledge Discovery Committee, CAAI Knowledge Engineering and Distributed Intelligence Committee, Professor of Kunming University of Science and Technology, Young Top Talent of Yunnan Ten Thousand Talents Program, Associate Editor of International Journal of Machine Learning and Cybernetics, deputy director of the Data Science Research Center of Kunming University of Science and Technology, shared the report of "Conceptual Cognitive Learning Based on Granular Computing". Taking granular computing as the basic idea and formal concept analysis as the main tool, he elaborated on the mathematical model and classification performance of concept cognitive learning. He raised some scientific questions about the impact of incremental learning and concept forgetting on classification performance.
Zhang Junbo, director of the Artificial Intelligence Laboratory of JD Smart City Research Institute and head of JD's Urban AI Product Department, took "Spatiotemporal AI and Urban Computing" as the theme. From the perspective of the application of artificial intelligence in the industry, he proposed the closed-loop of artificial intelligence technology innovation and application, that is, demand traction, technology research and development, product building, and finally application landing, by using the spatiotemporal machine learning and spatiotemporal deep learning of AI and big data to build an intelligent city, which had a good guiding significance for the innovation and application exploration of artificial intelligence technology.
The panel was presided over by Prof. Mi Jusheng.The forum chairman Prof. Miao Chaoqian, keynote speakers Gong Zhiguo, Hu Qinghua, Qian Yuhua, Yue Xiaodong and Li Jinhai participated in the discussion. The guests discussed the following questions and answered online questions from audiences, such as what are the uncertainties of uncertain AI? Do they cause trouble or bring good news to AI? What are the characteristics of granular computing in depicting uncertain information? What are the means to deal with uncertainty? What are the future research directions of granular computing? What are their fusion or support functions in other fields of AI? What are the advantages of granular computing in dealing with classification problems? The views of the participants provided new ideas for the scientific research of granular computing and uncertainty in the field of artificial intelligence.