Improved upper bounds on the expected error in constant step-size Q-learning.
Carolyn L. BeckR. SrikantPublished in: ACC (2013)
Keyphrases
- step size
- upper bound
- lower bound
- convergence rate
- expected error
- cost function
- convergence speed
- learning rate
- worst case
- reinforcement learning
- lower and upper bounds
- upper and lower bounds
- state space
- learning algorithm
- generalization error
- vc dimension
- np hard
- sample complexity
- gradient method
- sample size
- simulated annealing
- multiresolution
- image processing