Policy and Value Transfer in Lifelong Reinforcement Learning.
David AbelYuu JinnaiSophie Yue GuoGeorge Dimitri KonidarisMichael L. LittmanPublished in: ICML (2018)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- transfer learning
- action selection
- markov decision process
- reinforcement learning problems
- control policies
- function approximators
- control policy
- reward function
- partially observable environments
- function approximation
- policy gradient
- state space
- markov decision processes
- action space
- actor critic
- policy evaluation
- reinforcement learning algorithms
- policy iteration
- state action
- state and action spaces
- partially observable
- rl algorithms
- average reward
- markov decision problems
- partially observable markov decision processes
- lifelong learning
- dynamic programming
- temporal difference
- model free
- decision problems
- infinite horizon
- multi agent
- knowledge transfer
- approximate dynamic programming
- previously learned
- continuous state
- model free reinforcement learning
- control problems
- continuous state spaces
- supervised learning
- policy making
- transferring knowledge
- learning scenarios
- temporal difference learning
- optimal control
- cross domain
- long run
- average cost
- machine learning
- technology enhanced learning
- learning problems
- learning activities
- learning algorithm