大会报告


  • 大会报告 1

    戴琼海

    中国工程院院士

    中国人工智能学会理事长

    清华大学信息科学技术学院院长

    清华大学脑与认知科学研究院院长


    报告题目:人工智能---交互

    报告摘要:人类社会的发展历经工业革命、电气革命和信息革命,已正式跨入人工智能时代。人工智能的核心任务是赋能人类,并与人类和谐发展。因此,人工智能与人类,以及与物理世界的交互是真正提高人类认知和改造世界能力的关键,并可改变未来社会形态和服务模式。报告围绕上述交互问题,针对未来两大核心交互技术:与人类之间的增强现实交互,以及与物理世界之间的具身交互,阐释其技术挑战、发展机遇,展示了若干颠覆性用户体验案例,并展望了如何推动人类认知、发展与文化艺术变革的前景。

  • 大会报告 2

    陈俊龙

    华南理工大学特聘讲席教授

    欧洲科学院外籍院士

    中国自动化学会副理事长

    IEEE Fellow、美国科学促进会AAAS Fellow


    报告题目:Recurrent and Gated Broad Learning Systems: an LSTM-Like Architecture and its Applicationsin Simultaneous Learning of Time-Related Information

    报告摘要:In this talk, the broad learning system(BLS) will be briefly discussed. Then its variations in recurrent structure, Recurrent BLS (RBLS) and long-shorttermmemory (LSTM)-like architectures by adding forget gates function, Gated BLS (GBLS), will be discussed along with the learning algorithms. Recurrent BLS and Gated BLS and associated learning algorithms possess three advantages: 1) higher accuracydue to the simultaneous learning of multiple information,even compared to deep LSTM that extracts deeper but singleinformation only; 2) significantly faster training time dueto the noniterative learning in BLS, compared to LSTM; and3) easy integration with other discriminant information for furtherimprovement. The proposed methods have been evaluatedover 13 real-world datasets from various types of text classification.Compared to RBLS, GBLS hasan extra forget gate to control the flow of information (similarto LSTM) to further improve the accuracy on text classification. From the experimental results, the proposed methodsachieve higher accuracies than that of LSTM while taking significantlyless training time on most evaluated datasets, especially when theLSTM is in deep architecture.

  • 大会报告 3

    Roman Słowiński

    欧洲科学院院士

    波兰科学院院士


    报告题目:基于稳健序数回归的偏好学习用于多属性决策辅助

    报告摘要:识别决策者(DM)的偏好对于多属性决策辅助至关重要。我们提出一种建设性的偏好学习方法,称为稳健序数回归。这种方法论将运筹学与人工智能联系起来,从而证实了OR和AI之间相互关系的当前趋势。

    讲座从以下观察开始:在对多个属性(标准,投票者或自然状态)进行评估的一组备选方案中建立的优势关系是源自制定多属性决策问题(常规分类,排名,或选择;多目标优化是一种特殊情况)的唯一客观信息。尽管它可以消除许多不相关(即占主导地位)的替代方案,但它却使许多替代方案无法比拟。可以通过考虑DM的首选项来解决这种情况。因此,决策辅助方法需要一些具有单个或多个DM的价值体系的首选项信息。此信息通常具有决策示例的形式。稳健的序数回归将其用于构建偏好模型,然后将该偏好模型应用于一组非主导的备选方案,以得出呈现给DM的推荐。在实际决策协助中,由偏好启发,偏好建模和DM对建议的分析组成的过程一直循环进行,直到DM(或一组DM)接受建议或决定更改问题设置。这种互动过程称为建设性偏好学习。

  • 大会报告 4

    Shuzhi Sam Ge

    新加坡工程院院士

    新加坡国立大学教授

    IEEE Fellow, IFAC Fellow, IET Fellow, IES Fellow.


    报告题目:Spiral Advancementof AI and Robotics

    报告摘要:Robotics and Artificial Intelligence (AI) technologies have been widely employed in our works and daily lives as they are revolutionizing. For new-generation robot, one focus is the ability to have harmonious and natural social interaction with human. This is considered as one of the important signs that human society have entered the era of man-machine integration. In this lecture, I will firstly give a brief introduction about development of robot-human interaction technologies in short-term and specific scenarios, then go through the improvement that AI technologies bring up. I will focus more on the recent research works and advances in the social robotics based on knowledge learning, intelligent control, scene understanding with AI, companionship, and among others. I would like to conclude my lecture by the discussion on comprehensive robot interaction theory that integrates social knowledge, intention understanding, behavior generation and delivery. For each of the subtopics, the materials are prepared with pictures, and followed by some fundamental technical details. At this point of the time, the fusion of AI and robotics is apparent, and open for us to work together to make social robots as an integral part of our social fabric.

  • 大会报告 5

    孙富春

    清华大学教授

    中国人工智能学会副理事长

    国家杰青

    IEEE Fellow


    报告题目:Behavior-based Artificial Intelligence:Model and Approaches

    报告摘要:In the development of human society, the invention and use of machine is an important sign of human being's entering into industrialization, and machine becomes an independent actuator, which realizes the separation of executor and action instruction sender from our human. Furthermore, the emergence of artificial intelligence makes it possible for a machine to become a sender of instructions through interactions with environment and human, and as a result, the sender and executor of the instruction are unified on the machine itself. This report systematically analyzes behavior-based artificial intelligence from evolution of automatic control. Then, taking "how to sense like a human being" as the topic, the talk puts forward the framework of robot active perception, introduces the main achievements of the team in visual multi-target detection, visual tactile representation, multimodal fusion and developmental learning, and takes "how to operate like a human being" as the problem, gives the main achievements of the team in learning smart operation skills such as active imitation learning and preference learning achievements. Next, a new concept “Bcent” is proposed to model the unity of learning and coordination among sensing, cognition and behavior, and a new theoretical framework for behavior learning is proposed yet. Finally, the challenge and problems for behavior AI are discussed.

  • 大会报告 6

    岳东

    南京邮电大学教授,自动化学院、人工智能学院院长

    教育部长江学者特聘教授


    报告题目:信物融合有源配电网的协同智能控制与主动安全防御研究

    报告摘要:首先介绍配电网发展过程中面临的控制与信物安全的挑战性问题,进一步介绍在协同智能控制方面的研究成果,以及面向信物安全的主动安全防御研究思路和初步成果。最后,提出若干值得进一步研究的问题。

  • 大会报告 7

    徐昕

    国防科技大学教授

    国家杰青


    报告题目:Advances in Autonomous Control and Hybrid Intelligence of Unmanned Systems

    报告摘要:This talk analyzes the technical requirements and challenges of intelligent unmanned systems. Some advances in bio-inspired autonomous perception of complex environments and reinforcement learning for optimized decision-making and motion control are introduced. The learning-based cooperative control scheme of human-machine systems will be presented with some recent progresses. At last, the directions for further research and development in related areas are discussed.