Enhanced Meta Reinforcement Learning using Demonstrations in Sparse Reward Environments.
Desik RengarajanSapana ChaudharyJaewon KimDileep KalathilSrinivas ShakkottaiPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- function approximation
- reinforcement learning algorithms
- eligibility traces
- state space
- model free
- reward function
- dynamic environments
- optimal policy
- temporal difference
- markov decision processes
- partially observable environments
- multi agent environments
- average reward
- learning algorithm
- high dimensional
- sparse data
- learning agent
- policy iteration
- data sets
- robotic control
- machine learning
- reinforcement learning methods
- function approximators
- meta level
- partially observable
- learning process
- supervised learning
- transfer learning
- action selection
- learning problems