On tight bounds for function approximation error in risk-sensitive reinforcement learning.
Prasenjit KarmakarShalabh BhatnagarPublished in: Syst. Control. Lett. (2021)
Keyphrases
- approximation error
- tight bounds
- risk sensitive
- reinforcement learning
- optimal control
- markov decision processes
- model free
- upper bound
- utility function
- machine learning
- optimality criterion
- function approximation
- function approximators
- estimation error
- learning algorithm
- markov decision problems
- optimal policy
- state space
- temporal difference
- expected utility
- control policies
- dynamic programming
- cost function
- reconstruction error
- control strategies
- reinforcement learning algorithms
- partially observable
- policy iteration
- image classification
- decision making