KAIST Professor Jongse Park's Research Team Wins ISCA 'Best Research Artifact Award'

KAIST announced on the 1st that Professor Jongse Park's research team from the School of Computing recently won the Best Research Artifact Award at the ‘2024 International Symposium on Computer Architecture (ISCA)’ held in Buenos Aires, Argentina, for their paper titled ‘Accelerated Continual Learning Techniques for Video Analysis in Autonomous Systems.’

Diagram of continuous learning computation workflow on autonomous systems. Provided by KAIST

ISCA is the most prestigious international conference in the field of computer architecture. This year, a total of 423 papers were submitted to ISCA, of which 83 were accepted (acceptance rate 19.6%).

The Best Research Artifact Award is selected by ISCA based on the innovation, usability, and impact of the research artifacts.

The research team was positively evaluated for being the first to develop a neural processing unit (NPU) architecture and on-device software system to accelerate ‘continual learning,’ a foundational technology for adaptive artificial intelligence, as well as for the high completeness of their open-sourced code and data.

The main achievement of the research team is realizing ‘adaptive artificial intelligence’ using only on-device resources through AI chips embedded in devices.

The research results are expected to be applied across multiple fields, including building on-device AI systems in future mobility environments represented by Software-Defined Vehicles and Software-Defined Robots.

(From left) Professor Jongse Park, Ph.D. candidate Yunsung Kim, Ph.D. candidate Changhoon Oh, Department of Computer Science, KAIST. Provided by KAIST

Professor Jongse Park said, “It is meaningful to have demonstrated that adaptive artificial intelligence can be realized using only on-device resources through this research,” and added, “We will continue to pursue hardware and software research for on-device artificial intelligence.”

Meanwhile, this research was conducted with support from the National Research Foundation of Korea’s Excellent Young Researcher Support Program, the Institute for Information & Communications Technology Planning & Evaluation (IITP), the University ICT Research Center (ITRC), the Graduate School of Artificial Intelligence Support Program, and the Graduate School of AI Semiconductor Support Program.

The research team included KAIST School of Computing students Yunseong Kim, Changhoon Oh, Jinwoo Hwang, Wonwoong Kim, Sungryong Oh, and Yubin Lee; Dr. Hardik Sharma from Meta; Dr. Amir Yazdanbakhsh from Google DeepMind; and Professor Jongse Park from KAIST School of Computing.

This content was produced with the assistance of AI translation services.