Risk-Sensitive RL Using Sampling-Based Expectation-Maximization.
Erfaun NooraniJohn S. BarasKarl Henrik JohanssonPublished in: CDC (2023)
Keyphrases
- risk sensitive
- expectation maximization
- model free
- markov decision processes
- reinforcement learning
- optimal control
- em algorithm
- control policies
- maximum likelihood
- markov decision problems
- probabilistic model
- reinforcement learning algorithms
- utility function
- optimal policy
- function approximation
- generative model
- markov decision chains
- monte carlo
- markov decision process
- temporal difference
- policy iteration
- action space
- state space
- decision theoretic
- control strategies
- dynamic programming
- finite state
- control policy
- average reward
- image segmentation
- hidden markov models
- optimality criterion
- real valued
- multi agent
- machine learning