Welcome to Deep Representation Learning Research Group.

Our main research goal is to develop a better understanding on how deep learning works, especially from the view point of representation learning. Our research scope covers a variety of topics in deep learning including unsupervised learning, information theoretic approaches, representation learning, representations of transformers, interpretation of representations, meta-learning, NAS, structured compression of DNN, etc.

(For potential students in Korea: 현재는 Deep Learning, Representation Learning 관련 연구만 진행하고 있습니다 (Representation as Encoding, Meta-Learning, Self-supervised Learning, Information Theory in Deep Learning, BERT on IR, DNN compression, NAS, DNN Generalization Bounds). 수학과 프로그래밍에 충분히 준비가 되셨거나 성장가능성이 충분하신 경우에 지원하시기 바랍니다. *연구분야소개영상)