On the afternoon of July 11, the"HUST-NUS" Academic Lecture Series of the School of Computer Science and Technology was held online. Professor Bingsheng He from the School of Computing of the National University of Singapore(NUS)was invited to talk about Federated Learning Systems: Towards Effective and Efficient Machine Learning Systems on Data Silos. The lecture was hosted by Xuanhua Shi, Deputy Dean of the School of Computer Science and Technology, Huazhong University of Science and Technology(HUST), attracting more than 1200 researchers and students from NUS, HUST, and other universities at home and abroad.

In Professor He's opinion, the current Internet economy is developing rapidly, and data privacy is getting more and more attention from all parties based on the closed-loop thinking of the Internet economy. When the data collection capacity is limited or even unavailable, the closed loop of the Internet economy is easy to be broken. On this basis, How to deal with and develop the performance and scalability of computer systems is worth thinking about and studying intensely.

First, Professor He briefly introduced the origin and motivation of the federated learning system, and how to make our system more standardized under the supervision of data. According to He, the federated learning system has become a research hotspot because it can achieve collaborative training of machine learning models between different organizations under privacy restrictions.

Then, Professor introduced the research progress of federated learning and the future direction of systematic research. He compared the existing federated learning systems with vivid examples and introduced his team's research content in this direction in a simple way, including the overall framework of the model and some specific research methods. Inspired by the federated systems in other fields such as database and cloud computing, Professor He and his team studied the system design requirements of federated learning systems, and found that the heterogeneity and autonomy of federated learning systems in other fields were rarely considered in the existing federated learning systems, and further analyzed and forecast the research direction of federated learning systems in the future.

In the last discussion, Professor He patiently answered students' questions and had an in-depth exchange and discussion with the students on the content and subject issues of the lecture. This symposium not only provided opportunities for international academic exchanges between teachers and students but also provided new ideas and directions for research in related fields. It is also conducive to further cooperation between HUST and NUS in the future.