Fast active learning for pure exploration in reinforcement learning.
Pierre MénardOmar Darwiche DominguesAnders JonssonEmilie KaufmannEdouard LeurentMichal ValkoPublished in: CoRR (2020)
Keyphrases
- active learning
- exploration exploitation
- active exploration
- reinforcement learning
- exploration strategy
- learning algorithm
- training examples
- bandit problems
- supervised learning
- exploration exploitation tradeoff
- transfer learning
- machine learning
- training set
- state space
- selective sampling
- reinforcement learning algorithms
- learning process
- learning strategies
- random sampling
- labeled data
- experimental design
- data sets
- genetic algorithm
- action selection
- function approximation
- cost sensitive
- multi agent
- relevance feedback
- learning problems
- markov decision processes
- training data
- unlabeled data
- temporal difference
- semi supervised learning
- temporal difference learning
- support vector
- feature selection
- active learning strategies
- mobile robot
- neural network
- robotic control
- dynamic programming