CSE DSI Machine Learning Seminar with Xinlei Chen (FAIR, Meta AI)

Three Episodes of Ablative Research

In this talk, Chen will mainly share three pieces of his work driven by the philosophy of “ablative research”. Unlike research that enhances state-of-the-art by introducing well-established and well-proven techniques, he explores the removal of components that are previously thought as essential from the pipeline. The research he is sharing covers multi-modal image-and-language understanding, self-supervised learning, and transfer learning of pre-trained models, suggesting the broad applicability of this meta-level approach for scientific research.

Xinlei Chen is a research scientist at FAIR, Meta AI. His research interests are in computer vision and machine learning. He got his Ph.D. from Carnegie Mellon University in 2018, and a B.S. from Zhejiang University in 2012. He is a recipient of CVPR best paper honorable mention and ICML outstanding paper honorable mention awards in 2021, and two CVPR 2022 best paper finalists, all of which on self-supervised representation learning.