* 이번 학기는 온라인으로 열리게 되며 학부생, 대학원생, 교직원 누구나 참여 가능합니다.
세미나 2 – 정보통신대학원 세미나 2021년 1학기
Data Efficiency and Privacy Preservation for Personalized Machine Learning Models: from the Perspective of Audio Applications
김민제 교수 (Indiana University)
초록
Recent advancements in AI rely much on the availability and quality of training datasets. While crowdsourcing has streamlined the labeling process involving human annotators, it has been limited to certain classification tasks, such as ImageNet classification. On the other hand, in many other real-world AI problems, data efficiency is key to success: the labeled dataset is either too expensive to acquire or doesn’t even exist. This talk introduces the perspective of audio-related applications to the data efficiency issue. First, we will see an AI system that converts a bad singing voice into an in-tune version in a karaoke setup. Using this application, we will learn how the dataset’s labeling quality affects the performance. Then, personalized machine learning models are introduced as a solution to improving the performance and efficiency of an ML system. Finally, I will introduce carefully designed data-efficient learning methods to train the personalized models without sacrificing the privacy of the users.
Bio: Minje Kim is an assistant professor in the Dept. of Intelligent Systems Engineering at Indiana University, where he leads his research group, Signals and AI Group in Engineering (SAIGE). He is an Amazon Visiting Academic, consulting for Amazon Lab126. At IU, he is also affiliated with Data Science Program, Cognitive Science Program, Dept. of Statistics, and Center for Machine Learning. He earned his Ph.D. in Computer Science at the University of Illinois at Urbana-Champaign, MS in CSE at POSTECH, and BE in Information & Computer Engineering at Ajou Univ. He worked as a researcher at ETRI from 2006 to 2011. He received 2020 IEEE SPS Best Paper Award and Google and Starkey grants also honored his ICASSP papers as an outstanding student paper in 2013 and 2014, respectively. He is an IEEE Senior Member and also an elected member of the IEEE Audio and Acoustic Signal Processing Technical Committee (2018-2023). He is serving as an Associate Editor for EURASIP Journal and IEEE Open Journal of Signal Processing. He also serves as a reviewer, program committee member, or area chair for NeurIPS, ICML, AAAI, IJCAI, ICLR, ICASSP, Interspeech, ISMIR, IEEE T-ASLP, etc. He is on more than 50 patents as an inventor.
– 일시 3월 31일 (수), 13시
– zoom주소 : 학부공지 4878번 참고
– 회의 ID 862 4495 4271
– 문의 4단계 BK21 AI 사업단 (260-3150)
