AI Thrust Seminar | Towards Sample-efficient Overparameterized Meta-learning
Meta-learning typically involves two phases. First, one learns a suitable representation from the previously seen tasks. Secondly, this representation is used for learning a new task using only a few samples (i.e., few-shot learning). In this talk I will discuss:
1. Sample complexity of representation learning with general covariance,
2. Algorithm & analysis for overparameterized few-shot learning,
3. End to end performance of linear meta-learning.
Yue Sun is currently a software engineer at Microsoft, US. He got his Ph.D degree in Electrical Engineering from University of Washington, Seattle, US, and his bachelor degree in Electrical Engineering from Tsinghua University, China. He held research positions in Google and Nokia Bell Labs. His research was about theoretical understanding of optimization, ML and control, with projects on theoretical analysis of convergence of nonconvex optimization and statistical guarantee of structured models.