Considering Unseen States as Impossible in Factored Reinforcement Learning.
Olga KozlovaOlivier SigaudPierre-Henri WuilleminChristophe MeyerPublished in: ECML/PKDD (1) (2009)
Keyphrases
- reinforcement learning
- state space
- perceptual aliasing
- factored markov decision processes
- state abstraction
- state variables
- function approximation
- markov decision problems
- belief state
- state action
- transition model
- initial state
- reinforcement learning algorithms
- state transition
- optimal policy
- fully observable
- previously unseen
- markov decision processes
- hidden state
- training set
- multi agent
- learning algorithm
- approximate dynamic programming
- action selection
- markov decision process
- temporal difference
- neural network
- finite state machines
- dynamic programming
- hidden markov models
- learning process
- machine learning
- temporal difference learning
- state information
- model free
- learning classifier systems
- dynamical systems
- policy search
- robotic control