VPE: Variational Policy Embedding for Transfer Reinforcement Learning.
Isac ArnekvistDanica KragicJohannes A. StorkPublished in: CoRR (2018)
Keyphrases
- reinforcement learning
- optimal policy
- transfer learning
- policy search
- action selection
- markov decision process
- partially observable environments
- actor critic
- reward function
- markov decision processes
- policy gradient
- state space
- markov decision problems
- policy iteration
- action space
- reinforcement learning problems
- partially observable
- control policy
- function approximation
- dynamic programming
- function approximators
- reinforcement learning algorithms
- state and action spaces
- policy evaluation
- state action
- model free
- control policies
- learning algorithm
- infinite horizon
- decision problems
- machine learning
- average reward
- rl algorithms
- partially observable markov decision processes
- approximate dynamic programming
- variational methods
- state dependent
- cross domain
- knowledge transfer
- agent learns
- continuous state spaces
- agent receives
- inverse reinforcement learning
- learning process
- optical flow
- vector space
- optimal control
- information hiding
- long run
- policy gradient methods
- transition model
- control problems
- learning problems
- learning tasks
- multi agent