[행사/세미나] 인공지능대학원 전문가 초청 세미나(Dr. Saehoon Kim, Research Scientist at Kakao Brain, 8/2(월), 14:00))
- 인공지능학과
- 조회수25172
- 2021-07-19
■ Title: Recent advances of self-supervised visual representation learning
■Speaker: Dr. Saehoon Kim, Research scientist @ Kakao Brain
■Time : 2021 August 2nd 14:00 ~ 16:00
■Location: Webex room: https://skku-ict.webex.com/skku-ict/j.php?MTID=m33c3de95776383eda379ac2c96b8c9f7
■ Private meeting:
After the talk, we provide a private meeting (one-on-one or one-to-many) with Dr. Kim. Please write your name and email address (only school email) if you want to meet him personally until July 31th.
https://forms.gle/T5JzNttFvV1Vgep16
■ Abstract:
Self-supervised learning (SSL) aims to learn semantically meaningful representations without having human annotations. Due to the advance of SSL algorithms, learning a representation from many unlabeled samples becomes very effective to improve the performance of various computer vision downstream tasks. In this talk, I will summarize the recent approaches of self-supervised visual representation learning in three categories: (a) SSL with contrastive learning, (b) SSL without having negative samples, and (c) SSL with generative models. Then, I will describe our attempts to resolve limitations of current approaches. Firstly, I will introduce a novel SSL objective for localization downstream tasks, which aims to learn invariant representations from the same cropped region under augmentations of a given image. Lastly, I will introduce an approach to learn a robust representation by jointly optimizing contrastive and generative losses in a unified framework.
■ Bio:
Saehoon Kim is a research scientist at Kakao Brain. He received his B.S and Ph.D. degrees from the Department of Computer Science and Engineering at POSTECH, in 2009 and 2018, respectively. After completing his Ph.D., he worked at AITRICS, an AI healthcare startup, as a senior research scientist. His research interests lie in generative models, self-supervised representation learning, and machine learning for healthcare. He received the best paper runner-up award at WSDM’15. He served as a senior program committee member for AAAI-2021 and IJCAI-2021.