Reducing Sample Complexity in Reinforcement Learning by Transferring Transition and Reward Probabilities.
Kouta OguniKazuyuki NarisawaAyumi ShinoharaPublished in: ICAART (1) (2014)
Keyphrases
- sample complexity
- reinforcement learning
- learning problems
- learning algorithm
- supervised learning
- transfer learning
- theoretical analysis
- pac learning
- sequential decision problems
- vc dimension
- upper bound
- reinforcement learning algorithms
- active learning
- state space
- special case
- markov decision processes
- machine learning
- lower bound
- generalization error
- training examples
- multi agent
- reward function
- concept classes
- irrelevant features
- number of irrelevant features
- learning tasks
- unsupervised learning
- sample size
- learning process
- average case
- partially observable
- training data
- long run
- semi supervised learning
- learning environment
- training samples
- optimal policy