to top



The Third SFU-ZJU Joint Symposium Comes to a Successful Conclusion


From May 26th to 27th, 2018, the 3rd ZJU-SFU Joint Symposium organized by Simon Fraser University (SFU) and Zhejiang University (ZJU) was held in CaoGuangBiao Building at Yuquan Campus. Prof. Wu Fei, Vice Dean of ZJU College of Computer Science and Technology, extended a warm welcome to the participants and analyzed the significance of the symposium from the perspective of China's 15-year artificial intelligence strategic plan.

The symposium lasted two days. Guest scholars from domestic and foreign universities discussed cutting-edge technologies and real-world applications in the field of big data and visual computing.

At the symposium on May 26th, Prof. Richard Zhang from SFU first took a natural phenomenon, “symmetry is ubiquitous in nature and man-made artifacts”, as the entry point and argued that the ultimate goal of shape understanding is functional understanding. In addition, he shared his latest research results on functional analysis and modeling of three-dimensional shapes and environments, and also described new research directions.

Prof. Wu Fei from ZJU College of Computer Science and Technology focused on the application of memory-augmented learning in the media field, such as machine translation, visual conversation, image subtitle, and video actions, highlighting the superiority of this neural network. Memory-augmented learning is also an appropriate method to integrate data-driven learning, knowledge-guided inference and experience exploration.

Prof. Tan Ping from SFU gave a detailed introduction to the latest advancements in polarimetric 3D vision. This perspective combines per-pixel photometric information from polarization with epipolar constraints from multiple views for multi-view stereo and dense monocular SLAM. This approach solves the defect of unable to restore objects without feature areas when dealing with featureless scenes.

Prof. Robert D. Cameron from SFU pointed out that in the current technology environment, search performance per node is limited by the fundamentally sequential nature of search technologies such as grep, thus suggesting a search method for parallel execution of his research: Parabix Regular Expression Search. In the future, building a powerful compiler to solve the coding problem will be the next goal of Professor Robert D. Cameron.

Professor Sun Qi from ZJU College of Computer Science and Technology presented the research results of Prof. Wu Jian's team. He proposed that RealDoctor AI is a model for combining medical treatment with artificial intelligence. Its goal can be summarized as 1 + X + 1 + 1. Three ones represent medical AI conference, medical AI Club activities and cooperation between professors from universities or AI research centres with doctors in the hospitals. The X contains all aspects that medical AI can affect. The vision of RealDoctor AI is to promote and enhance medical services across the world.

Prof. Wang Jiannan from SFU discussed some of the challenges in the current state of data processing and introduced two systems that his lab is currently developing. The first system, DeepER, focuses on the data preparation phase, while the second system, AQP ++, focuses on the data exploration phase. Prof. Wang further explored future research directions in data processing.

Prof. Cai Deng from ZJU College of Computer Science and Technology pointed out in his speech that Nearest Neighbor Search (ANNS) is a basic problem in machine learning and data mining. Therefore, it is paramount for the ANNS algorithm to have high efficiency in memory usage and search performance. In the course of the study, Professor Cai Deng proposed a new graph structure named Navigation Spreading-out Graph (NSG), and a number of experiments showed that the algorithm is significantly better than all other existing algorithms.

With the rapid development of internet services, more information about the relationships between users in social media has also been discovered, opening new opportunities for recommendation systems. Dr. Wang Xin , a graduate in computer science from Tsinghua University (also a DDP undergraduate from the class of 2007 and a GDDP Ph.D. student from the class of 2011) led the research direction of the recommendation system in the new Internet era. He pointed out that a key task in this research is to model user preferences and to suggest a personalized list of items that user has not experienced.

Wang Beidou, a graduate student of the Ph.D. Dual Degree Program (GDDP) at ZJU College of Computer Science and Technology, focused on three prioritization related research questions on two popular types of broadcast messages, broadcast emails, and tweets. Rank feature, rank relevance and machine learning were taken into consideration in the first research; through analysis of tweet preference and influence of user’s decisions, a recommendation framework was established. In the second and third research, active learning and cross-domain recommendation technologies incorporate collaborative filtering into broadcast email priorities.

At the symposium on May 27th, Prof. Cao Nan from Tongji University talked about the tremendous progress made in anomaly detection in finance, internet, urban computing and healthcare, and reviewed developments in visual analysis technology. Prof. Cao Nan also presented his research project “Volia: Visual Anomaly Detection and Flow Spatio-Temporal Data Monitoring” and “ECGLens”, both of which are typical applications of anomaly detection technology.

Prof. Zhang Jun from Denver University and Wuhan University, first introduced the background and concept, definition, essence and coordination cognitive intelligence of the Internet Blueprint (IoM), and further described the key issues of the Internet of Things in artificial intelligence and platform technology. Prof. Zhang provided detailed explanations for using big data to fill the gap between social and hardware cognition.

Prof. Greg Mori of SFU presented his work on the deep learning methods recently established by his research team and models for learning trajectory features that represent individual behavior, namely tag structure, time structure and group structure.

The research topic presented by Dr. Wu Sai from ZJU College of Computer Science and Technology is Optimization Techniques for Distributed Databases. Given that query optimization is very challenging in large-scale distributed database systems, Prof. Wu used machine learning methods to approach the problem, especially in the optimization section. Prof. Wu covered topics such as querying distributed representation, neural cost model for query processing, learning-based query optimizer, and overall database tuning architecture.

With the acceleration in the pace of big data and visual computing development, the fields of environment, medicine, engineering, business, social sciences and humanities are all facing new opportunities and challenges. The 3rd ZJU-SFU Joint Academic Symposium organized by Simon Fraser University and Zhejiang University provided a new platform for cooperation and exchange in research. Although the symposium has come to a close, new ideas and developments are bound to usher in the field of big data and visual computing.

The College of Computer Science and Technology educates future leaders in computer science with interdisciplinary innovation capabilities to address global challenges in the AI2.0 world.