Label-Efficient Deep Learning
Dr. Sohn is a research scientist at Google. Prior to that, he worked at NEC Labs as a researcher after completing his PhD in 2015 from the department of Electrical Engineering and Computer Science at the University of Michigan. His research focuses on data efficient learning including semi-and self-supervised learning, domain adaptation, one-class classification, and deep generative models.
Machine learning for industrial applications is often constrained by the lack of large-scale annotated data, motivating research on label efficient learning. In this talk I’ll discuss recent progress on label efficient deep learning, including semi-supervised and self-supervised learning. Firstly, I’ll discuss consistency-based self-training methods for semi-supervised learning and domain adaptation of object classification and detection. Secondly, I’ll talk about novel applications of self-supervised representation learning for visual anomaly detection.