junhyuk (at) google (dot) com

Curriculum Vitae     Google Scholar

I am a research scientist at DeepMind. I finished my Ph.D. at the University of Michigan, where I was co-advised by Honglak Lee and Satinder Singh. My research focuses on deep reinforcement learning. Some of my work has been covered by MIT Technology Review and Daily Mail.

News

• June 2018: I joined DeepMind as a research scientist.
• May 2018: I will co-organize ICML 2018 Exploration in RL Workshop.
• Sep 2017: I will co-organize NIPS 2017 Deep Reinforcement Learning Symposium.
• June 2017: I will intern at DeepMind, advised by David Silver (Summer 2017).
• June 2016: I will intern at Microsoft Research, Redmond, advised by Pushmeet Kohli (Summer 2016).

Preprints

Contingency-Aware Exploration in Reinforcement Learning

Jongwook Choi*, Yijie Guo*, Marcin Moczulski*, Junhyuk Oh, Neal Wu, Mohammad Norouzi, Honglak Lee.

ArXiv, 2018. 

[Paper]

Many-Goals Reinforcement Learning

Vivek Veeriah, Junhyuk Oh, Satinder Singh

ArXiv, 2018. 

[Paper]

        

Unicorn: Continual Learning with a Universal, Off-policy Agent

Daniel J. Mankowitz, Augustin Žídek, André Barreto, Dan Horgan, Matteo Hessel, John Quan, Junhyuk Oh, Hado van Hasselt, David Silver, Tom Schaul

ArXiv, 2018. 

[Paper]

Publications

On Learning Intrinsic Rewards for Policy Gradient Methods

Zeyu Zheng, Junhyuk Oh, Satinder Singh

In Neural Information Processing Systems (NeurIPS), 2018. 

[Paper] [Code]

Hierarchical Reinforcement Learning for Zero-shot Generalization with Subtask Dependencies

Sungryull Sohn, Junhyuk Oh, Honglak Lee

In Neural Information Processing Systems (NeurIPS), 2018. 

[Paper]

Self-Imitation Learning

Junhyuk Oh*, Yijie Guo*, Satinder Singh, Honglak Lee

In International Conference on Machine Learning (ICML), 2018.

[Paper] [Code]

Value Prediction Network

Junhyuk Oh, Satinder Singh, Honglak Lee

In Neural Information Processing Systems (NeurIPS), 2017. 

[Paper] [Code]

Zero-Shot Task Generalization with Multi-Task Deep Reinforcement Learning

Junhyuk Oh, Satinder Singh, Honglak Lee, Pushmeet Kohli

In International Conference on Machine Learning (ICML), 2017. 

[Paper] [Project Page]

Control of Memory, Active Perception, and Action in Minecraft

Junhyuk Oh, Valliappa Chockalingam, Satinder Singh, Honglak Lee

In International Conference on Machine Learning (ICML), 2016. 

Press Coverage: [MIT Technology Review] [Daily Mail]

[Paper] [Project Page] [Code] 

Learning Transferrable Knowledge for Semantic Segmentation

 

with Deep Convolutional Neural Network 

Seunghoon Hong, Junhyuk Oh, Bohyung Han, Honglak Lee 

In Computer Vision and Pattern Recognition (CVPR), 2016.

Spotlight presentation

[Paper] [Project Page] [Code]

Action-Conditional Video Prediction using Deep Networks in Atari Games
Junhyuk Oh, Xiaoxiao Guo, Honglak Lee, Richard Lewis, Satinder Singh
In Neural Information Processing Systems (NeurIPS), 2015. 

Spotlight presentation Acceptance rate: 4.5% (82/1858)

[Paper] [Project Page] [Code]