I am a PhD student at the University of Toronto advised by Jimmy Ba and Geoffrey Hinton. I am also a researcher at Google Brain and the Vector Institute. I completed my MRes in CSML at University College London advised by Tim Lillicrap and Karl Friston. My work is supported by the Vanier Scholarship.

My goal is to understand the computational principles behind human thinking by replicating them in complex environments. My current research focuses on general objectives for reinforcement learning, reinforcement learning through world models, and unsupervised learning. You can hear more about my research on the TalkRL Podcast.

If you have any questions, feel free to contact me at [email protected].

Selected research

Please see my Scholar profile for a full list including article links.

• 

  Mastering Atari with Discrete World Models

  D Hafner, T Lillicrap, M Norouzi, J Ba Preprint

  Paper Twitter

• 

  Action and Perception as Divergence Minimization

  D Hafner, P Ortega, J Ba, T Parr, K Friston, N Heess Preprint

  Paper Twitter Poster FAQs

• 

  Planning to Explore via Self-Supervised World Models

  R Sekar, O Rybkin, K Daniilidis, P Abbeel, D Hafner, D Pathak ICML 2020 (22%)

  Blog Paper Code Twitter Talk Website

• 

  Dream to Control: Learning Behaviors by Latent Imagination

  D Hafner, T Lillicrap, J Ba, M Norouzi ICLR 2020 (oral, 4%), DRLW 2019 (oral)

  Blog Paper Twitter Poster Videos Code

• 

  A Deep Learning Framework for Neuroscience

  Richards, Lillicrap, Beaudoi, Bengio, Bogacz, Christensen, Clopath, Costa, Berker, Ganguli, Gillon, Hafner, Kepecs, Kriegeskorte, Latham, Lindsay, Miller, Naud, Pack, Poirazi, Roelfsema, Sacramento, Saxe, Scellier, Schapiro, Senn, Wayne, Yamins, Zenke, Zylberberg, Therien, Kording Nature Neuroscience

  Paper Twitter

• 

  Noise Contrastive Priors for Functional Uncertainty

  D Hafner, D Tran, A Irpan, T Lillicrap, J Davidson UAI 2019 (26%)

  Paper Poster Code

• 

  Sample-Efficient Reinforcement Learning with Stochastic Ensemble Value Expansion

  J Buckman, D Hafner, G Tucker, E Brevdo, H Lee NeurIPS 2018 (oral, 0.6%)

  Paper Poster Talk Code

• 

  Learning Latent Dynamics for Planning from Pixels

  D Hafner, T Lillicrap, I Fischer, R Villegas, D Ha, H Lee, J Davidson ICML 2019 (oral, 23%)

  Blog Paper Poster Code

• 

  Sim-to-Real: Learning Agile Locomotion For Quadruped Robots

  J Tan, T Zhang, E Coumans, A Iscen, Y Bai, D Hafner, S Bohez, V Vanhoucke RSS 2018 (oral, 31%)

  Paper

• 

  TensorFlow Agents: Efficient Batched Reinforcement Learning in TensorFlow

  D Hafner, J Davidson, V Vanhoucke Tech report

  Paper Code

• 

  Learning Hierarchical Information Flow with Recurrent Neural Modules

  D Hafner, A Irpan, J Davidson, N Heess NIPS 2017 (21%)

  Paper Poster

• 

  Deep Reinforcement Learning from Raw Pixels in Doom

  D Hafner BSc thesis

  Thesis Code

Short biography

Danijar Hafner is a PhD student at the University of Toronto advised by Jimmy Ba and Geoffrey Hinton. He is also a student researcher at Google Brain and the Vector Institute. His research focuses on artificial intelligence, specifically reinforcement learning with world models and intrinsic objectives for agents that learn without rewards. Danijar obtained his MRes at the University College London and the Gatsby Unit under the supervision of Tim Lillicrap and Karl Friston. He completed his bachelor’s thesis on deep reinforcement learning for the video game Doom at the Hasso Plattner Institute, Germany.

Links

Twitter · Github · Quora · LinkedIn · Google AI · ORCID