Option Discovery in Reinforcement Learning using Frequent Common Subsequences of Actions.
Sertan GirginFaruk PolatPublished in: CIMCA/IAWTIC (2005)
Keyphrases
- reinforcement learning
- perceptual aliasing
- action selection
- action space
- partially observable
- multiagent reinforcement learning
- frequent pattern discovery
- state and action spaces
- partially observable domains
- state space
- efficient discovery
- state action
- reward function
- function approximation
- machine learning
- reinforcement learning algorithms
- human activities
- knowledge discovery
- partial observability
- markov decision processes
- dynamic time warping
- learning algorithm
- multi agent
- frequent patterns
- initially unknown
- mining frequent
- learned knowledge
- transfer learning
- discovery process
- human actions
- pattern discovery
- state transitions
- learning agent
- partially observable markov decision process
- decision problems
- optimal policy
- sensing actions
- belief state
- temporal difference
- data mining