Fast active learning for pure exploration in reinforcement learning.
Pierre MénardOmar Darwiche DominguesAnders JonssonEmilie KaufmannEdouard LeurentMichal ValkoPublished in: ICML (2021)
Keyphrases
- active learning
- exploration exploitation
- active exploration
- reinforcement learning
- random sampling
- learning process
- learning algorithm
- machine learning
- exploration strategy
- semi supervised
- transfer learning
- relevance feedback
- model based reinforcement learning
- training examples
- bandit problems
- supervised learning
- training set
- state space
- action selection
- exploration exploitation tradeoff
- selective sampling
- pool based active learning
- model free
- markov decision processes
- optimal policy
- semi supervised learning
- experimental design
- reinforcement learning algorithms
- batch mode
- function approximation
- multi agent
- temporal difference
- autonomous learning
- sample selection
- learning strategies
- robotic control
- learning environment
- decision trees