摘要：

Domain generalization (DG) is a crucial problem. Researchers in this area are trying to navigate a way to improve models generalization ability and make them function well in different environments. To be specific, it aims to learn robust models from multiple source domains that could generalize well on unseen target domains. In this talk, the speaker will first introduce DG problem and then discuss four types of methods in this direction, including (1) Data Manipulation based methods  (2) Meta Learning based methods (3) Invariant Learning based methods (4) Ensemble Learning based methods.

简介：

Bo Li is a Phd student at NTU, Singapore. He studies computer vision and machine learning problem

with Prof. Ziwei Liu. His research papers can be found in https://scholar.google.com/citations?user=1_zc1-IAAAAJ&hl=en

摘要：

Domain generalization (DG) is a crucial problem. Researchers in this area are trying to navigate a way to improve models generalization ability and make them function well in different environments. To be specific, it aims to learn robust models from multiple source domains that could generalize well on unseen target domains. In this talk, the speaker will first introduce DG problem and then discuss four types of methods in this direction, including (1) Data Manipulation based methods  (2) Meta Learning based methods (3) Invariant Learning based methods (4) Ensemble Learning based methods.

简介：

Bo Li is a Phd student at NTU, Singapore. He studies computer vision and machine learning problem

with Prof. Ziwei Liu. His research papers can be found in https://scholar.google.com/citations?user=1_zc1-IAAAAJ&hl=en