▲ Gyu Seon Kim, Integrated M.S./Ph.D. Student (left), and Professor Joongheon Kim (right), School of Electrical Engineering, Korea University

Gyu Seon Kim, an integrated M.S./Ph.D. student in the AI and Mobility Lab (Advisor: Professor Joongheon Kim) at the School of Electrical Engineering, Korea University, has developed a novel technology called Slim Federated Reinforcement Learning (SlimFRL). This technology significantly enhances the efficiency of proactive caching systems, which predict and store content in advance based on user demand.
 

Proactive caching refers to storing content on a user's device or a nearby server before the user requests it. While similar to how services like Netflix or YouTube recommend videos a user might like, proactive caching goes a step further by actually downloading the content in advance. This allows the system to deliver content instantly when requested, reducing loading times and alleviating network congestion.
 

A major limitation of previous approaches was the inflexibility of neural network sizes in response to changes in user mobility or user count. The proposed SlimFRL resolves this by automatically adjusting the neural network size based on the number of users. Furthermore, since only locally trained models are shared, user privacy is preserved.
 

Recognizing the significance of this achievement, the research has been accepted for publication in IEEE Transactions on Networking (TON)—one of the premier international journals in the field of communications. Gyu Seon Kim’s work focused on developing a federated and reinforcement learning-based caching algorithm to improve energy efficiency in future intelligent network systems.