KAIST wins ‘Best Record Award’ for continuous learning technology for on-device AI

Professor Park Jong-se’s research team presenting research results
[Provided by KAIST. Redistribution and DB prohibition],
,
, “Daejeon= Yonhap News) Reporter Park Joo-young = The ‘Continuous Learning Acceleration Technique for Video Analysis of Autonomous Systems’ paper by Professor Park Jong-se’s research team in the Department of Computer Science at the Korea Advanced Institute of Science and Technology (KAIST) received the ‘Distinguished Artifact Award’ at the ‘2024 International Symposium on Computer Architecture’ (ISCA 2024) held in Buenos Aires, Argentina, from June 29 to July 3,” KAIST announced on the 1st. “,
,
, ‘The International Symposium on Computer Architecture (ISCA) is a prestigious international conference in the field of computer architecture, with a total of 423 papers submitted this year, of which only 83 papers (acceptance rate of 19.6%) were accepted.’,
,
, ‘Among the accepted papers, the Best Research Artifact Award is given considering the innovation, applicability, and impact of the research artifact.’,
,
, ‘The research by Professor Park’s team is a technology that can realize adaptive artificial intelligence (AI) using only on-device resources.’,
,
, ‘On-device AI is a technology that allows AI to process computations within the device itself without the need for communication with a server, enabling faster processing compared to AI driven through the cloud, drawing attention as AI smartphone or AI PC technology.’,
,
, ‘However, there is a problem of decreased accuracy due to the use of lightweight AI models with limited computation and memory resources.’,
,
, ‘Research on machine learning, system, and continuous learning techniques is actively being conducted as an adaptive AI methodology to address this issue.’,
,
, ‘The team developed a neural processing unit (NPU) structure and on-device software for continuous learning systems that can effectively increase the accuracy of lightweight models using only on-device resources, without the need for remote computing resources.’,
,
, ‘It is expected to contribute to the establishment of on-device AI systems in future mobility environments such as software-defined vehicles (SDVs) and software-defined robots (SDRs).’,
,
, ‘Professor Park Jong-se expressed his impressions, saying, “This achievement was made possible by the dedicated efforts of the students and close collaboration with researchers from Google and Meta,” and “We will continue our research on hardware and software for on-device AI in the future.”‘,
,
, ‘ [email protected]’,
,