On May 5, 2021, sponsored by CAAI, jointly organized by Human-machine integration Intelligence Committee, Institution of Intelligent Robot, Fudan University and Institution of Brain and Cognitive Science of Tsinghua University, the first online forum of Human-machine Intelligence was held at nineteen o’clock. This series of online forum has attracted wide attention, with an accumulating viewer of more than 320 thousand times. 

Showing the development potential in the field of human-computer Intelligence

This forum, hosted by Kang Xiaoyang, the associate researcher of Institute of intelligent robot of Fudan University, is themed with brain-computer interface and artificial intelligence. Qiao Hui, the assistant professor of Department of Automation of Tsinghua University, and Guo Yuchen, the assistant researcher of National Institute of Information Science and Technology of Tsinghua University and Jia Fumin, the associate researcher of Institute of Brain-like Intelligence and Technology Research of Fudan University were invited to give a report on this forum. They brought the viewers excellent academic report on the live-broadcasting platform, each sharing the latest research edge and relative research experience and then answered the questions asked by the viewers.

The successful holding of this forum provides all scholars with an exchange platform, enabling the viewers to have a deep understanding of some of the front direction and development of brain-computer interface and artificial intelligence, and also know the infinite developing potential in the field of brain-computer fusion intelligence. Meanwhile, it also provides some valuable experience for the further series of forum. 

Three Youth Scientists sharing Academic Reports

Professor Qiao Hui firstly introduced the unsupervised model transition microscopic imaging technology. During the processing of microscopic imaging, it is usually unfeasible to achieve paired training data in many occasions because of high-speed dynamic features of biological activity, incompatibility of imaging models or irreversibility of sample preparation. In order to promote the application of deep learning of image transformation in biomedical imaging, it’s necessary to break the dependence on paired training data. This report mainly focuses on the recent research and application exploration of Professor Qiao Hui and his group in unsupervised model transition optical microscopic imaging.

Nextly, the researcher Guo Yuchen introduced the theory and application of efficient deep learning. She mainly shared such questions on her report as following:

Big data and strong computing power laid a solid foundation for deep learning, but must the deep learning rely on big data and strong computing power? Under the internet data long-tail effect, how to build a deep learning model for tail data with few data? Among the widely existing edge devices (such as brain, phones), how to deploy the complexed deep learning models with limited computing resources? Her report discussed that under the condition of small data and weak computing power, efficient deep learning can be realized through brain-inspired theory.

Finally, the associate researcher Jia Fumin explained the frequency-dependent closed-loop control of the deep brain in the treatment of Parkinson’s disease. The high frequency electrical stimulation of the subthalamic nucleus (above 130Hz) in the treatment of Parkinson’s disease can not only obviously improve the symptoms of tremor, muscle rigidity and bradykinesia, but reduce the movement fluctuation caused by taking drugs and extend the opening time, which has become the first choice as a surgical treatment for Parkinson’s disease. However, such treatment has a limited improvement to freezing gait and other midline symptoms, or even worse. The relatively low frequency (60Hz-80Hz) can improved such symptoms mentioned above, but the control effect of the main symptoms of Parkinson’s disease will be reduced. In the process of domestic production of brain pacemakers, combining the clinical practice and theory, Jia’s team firstly proposed a new therapy for subthalamic nucleus frequency conversion stimulation for the treatment of freezing gait in Parkinson’s disease, which has got certain clinical effect. However, this therapy still needs further improvement and even basic research. How to combine close-loop stimulation with artificial intelligence technology in the future will face more challenges and opportunities.

During the question-and-answer session after each presentation, the online viewers asked questions in their respective fields of interest. And then the host selected some of these questions giving them to the guests of the report for answering. They have an enthusiastic interact with the online viewers.

This article is provided by CAAI Professional Committee Human-computer Fusion Intelligence.