Loading...

Meta Reinforcement Learning for Domain Generalization

Riyahi Madvar, Maryam | 2022

126 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55379 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rohban, Mohammad Hossein
  7. Abstract:
  8. Deep reinforcement learning has achieved better cumulative rewards than humans in many environments like Atari. One drawback of these methods is their data inefficiency which makes training time-consuming, and in some cases having this amount of data is infeasible. Meta reinforcement learning can use past experiences to enable agents to adapt to new tasks faster and makes neural networks to train in a short amount of time.One of the methods in meta reinforcement learning is inferring tasks which helps exploitation policy to have good performance in new tasks. There’s a need to improve exploration policy as well as exploitation policy by gaining informative transitions about the new task. Improving exploration policy causes better performance for exploitation policy and increases convergence speed.In this article, an exploration policy is used that gains information about the difference between dynamics between different tasks which helps generalization among tasks. Using unsupervised reinforcement learning, an intrinsic reward that is not dependent on the agent’s environment is used for reaching distinct trajectories. The intrinsic reward is computed by calculating the variance of neural networks’ output in the ensemble model. The conducted experiments on the MuJoCo simulator show improvement in comparison with the previous method named PEARL
  9. Keywords:
  10. Reinforcement Learning ; Domain Generalization ; Metalearning ; Unsupervised Learning ; Meta Reinforcement Learning ; Deep Reinforcement Learning

 Digital Object List

 Bookmark

No TOC