2023년 10월 13일 중앙대학교 웰에이징데이터과학융합연구소에서 학술 세미나를 개최하였습니다.

서울 시립대학교 김윤영 교수님께서 "Exploiting and Exploring the Uncertainty of Neural Networks via Perturbing Input Space" 주제로 강연해 주셨습니다.

초록: 

Deep Learning models have shown outstanding performance in various tasks, but they require a large amount of datasets in the training process. However, labeling them requires high computational cost and human resources. Among various research directions, active learning, which selects informative data instances via acquisition function, has been proposed for an efficient resource utilization. In this seminar, a brief review on the development of AL will be introduced, followed by the recent methodologies that have been proposed to deal with the selection bias in the conventional AL algorithms. As a future work, AL is capable to be adopted for finetuning or enhancing the performance of large language models.