Events

CILVR Seminar: Towards Real-World Meta-Learning

Speaker: Hae Beom Lee

Location: On-Line
Videoconference link: https://nyu.zoom.us/j/97986595706

Date: Tuesday, April 19, 2022

In this job talk, I will introduce my previous efforts extending the scope of meta-learning towards real-world learning scenarios. We will see how to deal with realistic task distributions including class/task imbalance and distributional shift, discuss how to develop practical meta-knowledge applicable to arbitrary CNN architectures, and also how to scale up few-shot learning to many-shot and heterogeneous learning problems. For future research agenda, I propose to explore more challenging meta-learning problems from the perspective of other adjacent research areas such as natural language processing and reinforcement learning, especially focusing on how to efficiently and effectively adapt to out-of-distribution tasks under non-stationary environments. I emphasize the importance of developing sophisticated meta-knowledge structures and will introduce a few ideas in this direction.