Haonan Yu

Research Scientist
Baidu Research
CV / Google Scholar / GitHub @yu239

My research focus is learning to ground language in vision and control, with deep reinforcement learning as the primary tool, under weak or no supervision. My previous work (during PhD) was learning to ground language in realistic video data, but with much more expensive supervision. I am also generally interested in writing machine learning code.

I received a Ph.D. degree in Computer Engineering from Purdue University in 2016.


Publications


Softwares

PaddlePaddle - Baidu's open-source deep learning platform which I occasionally contribute to.
XWorld - A C++/Python simulator package for reinforcement learning research.
FLARE - A PyTorch reinforcement learning framework for embodied agents (work in progress)


Slides

proximal policy optimization (PPO) - Slides for the RL reading group @SVAIL (07/31/2018).
language grounding by deep reinforcement learning - Slides for the talk @GoogleResearch (07/20/2018).
n-step bootstrapping - Slides for the RL reading group @SVAIL (05/10/2018).