Skip to content

Instantly share code, notes, and snippets.

@ih4cku
Last active August 29, 2015 14:22
Show Gist options
  • Save ih4cku/5564b2b5cb27a241a82c to your computer and use it in GitHub Desktop.
Save ih4cku/5564b2b5cb27a241a82c to your computer and use it in GitHub Desktop.
DL&RL

DL-RL

Introduction:

  • Deep learning (DL) is a new research track within the field of machine learning, building architectures consisting of multiple layers of representations in order to learn high level abstractions.
  • Reinforcement learning (RL) is one of the most promising AI paradigms for the future development of autonomous robots. RL allows a robot to learn from trial-and-error interactions with its environment.

Capability complementary:

  • DL can be used as a strong perception model of the environment.
  • RL is used as the learning paradigm for controling robot's actions.
  • The combination of DL and RL allows a learning agent to control a system based only on visual inputs, using a deep neural network to extract relevant features from the images.

Deep Learning is used as the perception model for:

  • Robots vision, aural and natural language comprehension

to:

  • provide representation of the environment,
  • and producing control signals.

Merits:

  • End-to-end system.
  • No need of physical modeling.
  • Learning good representations.

References:

  1. Human-level control through deep reinforcement learning, Volodymyr Mnih et al. Nature 518, 529–533.
  2. Deep Learning for Detecting Robotic Grasps, Ian Lenz, Honglak Lee, Ashutosh Saxena. International Journal of Robotics Research (IJRR), 2014.
  3. Playing Atari with Deep Reinforcement Learning, Volodymyr Mnih et al. CoRR Vol. abs/1312.5602.
  4. End-to-End Training of Deep Visuomotor Policies, Sergey Levine et al. CoRR Vol. abs/1504.00702.
  5. Peter Vrancx, Deep reinforcement learning
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment