|
|||||||||||||||||
| Employment • 2023.06 ~ Present: Research Scientist, Nokia Bell Labs, Cambridge, UK • 2023.03 ~ 2023.05: Postdoctoral Researcher, Networking & Mobile Systems Lab, KAIST, Republic of Korea • 2022.08 ~ 2022.11: Research Intern, Google Research, NYC, USA • 2019.04 ~ 2019.06: Research Intern, Microsoft Research, Beijing, China • 2018.07 ~ 2018.09: Research Intern, Nokia Bell Labs, Cambridge, UK Education • 2017.09 ~ 2023.02: Ph.D. in Computer Science, KAIST, Republic of Korea • 2016.03 ~ 2017.08: M.S. in Computer Science, KAIST, Republic of Korea • 2012.03 ~ 2016.02: B.S. in Computer Science, Yonsei University, Republic of Korea |
|||||||||||||||||
| Deep neural networks (DNNs) excel in various fields, but their effectiveness often diminishes due to discrepancies between training and real-world data. Such distributional shifts limit DNNs' applicability in critical areas. Test-time adaptation (TTA) addresses this by adapting DNNs to new, unseen domains using only unlabeled test data, eliminating the need for additional data gathering or labeling. This talk will explore two key studies that tackle the real-world deployment of TTA, focusing on challenges like non-uniform test data and unexpected diversity and noise in test streams. | |||||||||||||||||
|
|
|||||||||||||||||
|
|||||||||||||||||
후원기관 광고영상
more