Meta Reinforcement Learning with Task Embedding and Shared Policy.
Lin LanZhenguo LiXiaohong GuanPinghui WangPublished in: IJCAI (2019)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- action selection
- control policy
- function approximation
- function approximators
- state space
- actor critic
- markov decision processes
- approximate dynamic programming
- reward function
- policy iteration
- policy gradient
- action space
- reinforcement learning algorithms
- reinforcement learning problems
- partially observable
- control policies
- learning algorithm
- partially observable markov decision processes
- markov decision problems
- model free
- infinite horizon
- rl algorithms
- partially observable environments
- average reward
- decision problems
- partially observable domains
- state and action spaces
- agent learns
- state action
- finite state
- vector space
- optimal control
- machine learning
- inverse reinforcement learning
- agent receives
- dynamic programming
- policy gradient methods
- policy evaluation
- continuous state
- long run
- supervised learning
- transition model
- reinforcement learning methods
- temporal difference learning
- meta level
- state dependent
- control problems
- nonlinear dimensionality reduction
- temporal difference
- average cost