Group 1 - The seventh International Academic Conference on Artificial Intelligence, Computer Science, and Information Processing (AICSIP 2025) was held in Hangzhou, Zhejiang on July 25 [1] - A paper titled "Explaining The Improvement Of Multi-Exit Structure Distillation Using Stage Training And Attention Integration" was presented by a data intelligence technology company, attracting significant attention from both academia and the tech industry [1] - The research addresses the performance decline and inefficiency issues that arise when original AI models are distilled into smaller, less hardware-intensive models for enterprise applications, proposing a technique that enhances knowledge transfer efficiency during the distillation process [1] Group 2 - The technology has been successfully applied to a cloud-based hydropower model platform, resulting in a 52% reduction in computational consumption and a 40% increase in inference speed for the established runoff time series prediction model compared to pre-distillation [1] - This advancement significantly lowers the hardware resource requirements for AI applications in the hydropower industry, providing strong support for enhancing the intelligence level of hydropower applications [1] - The conference was sponsored by the IEEE China Council and co-hosted by Hangzhou Normal University and Hubei Zhongke Geological and Environmental Technology Service Center, with all accepted papers to be included in the IEEE Xplore core database and submitted for EI Compendex and Scopus indexing, highlighting the academic value and technological forefront of the research [2]
数智科技大数据公司科研成果获IEEE国际学术会议收录
Jing Ji Wang·2025-07-31 06:38