image: Four papers from UNIST Artificial Intelligence Graduate School accepted for presentation at the 2024 Association for the Advancement of Artificial Intelligence (AAAI).
Credit: UNIST
Four research papers by UNIST researchers have been accepted for presentation at the 2024 Association for the Advancement of Artificial Intelligence (AAAI) conference, a prestigious international academic event in the field of artificial intelligence held annually. These groundbreaking achievements will be unveiled at the 38th annual regular conference, scheduled to take place in Vancouver, Canada for eight days starting on February 20, 2024.
During this regular conference, a record-breaking number of over 12,000 papers were submitted from around the globe, with only 2,342, approximately 23%, being selected for adoption. The remarkable accomplishments of UNIST researchers emanated from the teams led by Professor Jae-Young Sim, Kyungdon Joo, Seungyul Han, and Gi-soo Kim, respectively.
Professor Sim’s team has developed cutting-edge computer vision technology that harnesses virtual data for artificial intelligence (AI) learning. This innovation aims to facilitate the analysis of CCTV footage to automatically identify specific individuals, such as criminals and missing persons. By leveraging virtual data, the technology significantly reduces data labeling costs and mitigates issues related to portrait rights infringement.
Another noteworthy advancement is the AI technology introduced by Professor Joo’s team, enhancing user experiences in virtual reality and gaming. This technology ensures precise 3D interactions among virtual characters, enabling realistic movements like hugging and handshakes. By synchronizing virtual humans’ actions with real-world gestures, immersive interactions are seamlessly recreated on a 3D platform.
Professor Han’s team will present their pioneering work on multi-agent reinforcement learning technology. This crucial AI innovation not only benefits gaming but also extends to applications like autonomous driving and robotic platooning, necessitating seamless collaboration between machines. Their high-efficiency algorithm streamlines the exploration process in multi-agent reinforcement learning, significantly reducing the number of cases to be evaluated.
Lastly, Professor Kim’s team has devised a sophisticated “bandit algorithm” for online recommendations, catering to users’ diverse preferences in news, video content, and shopping. This algorithm excels at suggesting similar items by analyzing user preferences, showcasing its ability to comprehend intricate user tastes effectively.
Established in 2020, the UNIST Artificial Intelligence Graduate School has made remarkable strides by recruiting talented young faculty members. Despite its short history, the school achieved notable success last year with a total of 31 papers accepted by top-tier conferences on AI and computer vision, such as NeurIPS and CVPR.