Effects of Task Similarity on Policy Transfer with Selective Exploration in Reinforcement Learning.
Akshay NarayanTze-Yun LeongPublished in: AAMAS (2019)
Keyphrases
- reinforcement learning
- action selection
- optimal policy
- transfer learning
- exploration exploitation tradeoff
- policy search
- active exploration
- exploration strategy
- function approximation
- markov decision process
- reinforcement learning problems
- reinforcement learning algorithms
- state space
- partially observable environments
- action space
- similarity measure
- autonomous learning
- similarity metric
- temporal difference
- markov decision problems
- control policies
- model based reinforcement learning
- learning algorithm
- dynamic programming
- machine learning
- actor critic
- approximate dynamic programming
- markov decision processes
- control policy
- function approximators
- knowledge transfer
- reinforcement learning methods
- learning tasks
- similarity function
- partially observable
- euclidean distance
- policy gradient
- neural network
- distance function
- model free
- infinite horizon
- state and action spaces
- decision problems
- finite state
- continuous state
- previously learned
- partially observable markov decision processes
- sufficient conditions