Unifying PAC and Regret: Uniform PAC Bounds for Episodic Reinforcement Learning.
Christoph DannTor LattimoreEmma BrunskillPublished in: NIPS (2017)
Keyphrases
- upper bound
- vc dimension
- pac bayesian
- reinforcement learning
- lower bound
- mistake bound
- sample complexity
- worst case
- distribution free
- confidence bounds
- pac learning
- sample size
- function approximation
- theoretical analysis
- generalization bounds
- noise tolerant
- statistical queries
- concept class
- concept classes
- online learning
- perceptron algorithm
- learning algorithm
- average case
- upper and lower bounds
- regret bounds
- expert advice
- linear threshold
- uniform convergence
- covering numbers
- reinforcement learning algorithms
- multi agent
- learning theory
- error bounds
- learning problems
- optimal policy
- optimal solution
- state space