Simulation-based optimization of Markov reward processes

IEEE Transactions on Automatic Control

Volumn 46, Issue 2, 2001, Pages 191-209

  Marbach, Peter ^b   Tsitsiklis, John N ^a,c  

^a IEEE   (United States)

^b UNIVERSITY OF CAMBRIDGE   (United Kingdom)

^c MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

Author keywords

Markov reward processes; Simulation based optimization; Stochastic approximation

Indexed keywords

MARKOV REWARD PROCESSES; STOCHASTIC APPROXIMATION;

ALGORITHMS; APPROXIMATION THEORY; COMPUTER SIMULATION; DYNAMIC PROGRAMMING; POLYNOMIALS; STATE SPACE METHODS;

MARKOV PROCESSES;

EID: 0035249254     PISSN: 00189286     EISSN: None     Source Type: Journal    
DOI: 10.1109/9.905687     Document Type: Article

References (30)

1. Stochastic approximation with two time scales
2. Neuron-like elements that can solve difficult learning control problems
3. Perturbation realization, potentials, and sensitivity analysis of Markov processes
4. Stochastic optimization of regenerative systems using infinitesimal perturbation analysis
5. Algorithms for sensitivity analysis of Markov systems through potentials and perturbation realization
6. General results on the convergence of stochastic algorithms
7. Smoothed perturbation analysis derivative estimation for Markov chains
8. Gradient estimation for regenerative processes
9. Stochastic approximation for Monte Carlo optimization
10. Likelihood ratio gradient estimation: An overview
11. Importance sampling for stochastic simulation
12. Reinforcement learning algorithm for partially observable Markov decision problems
13. Actor-critic like learning algorithms for Markov decision processes
14. Actor-critic algorithms
15. A unified view of the IPA, SF, and LR gradient estimation techniques
16. Analysis of recursive stochastic algorithms
17. Simulation-based optimization of Markov reward processes: Implementation issues
18. Policy gradient methods for reinforcement learning with function approximation
19. Practical issues in temporal difference learning
20. Missing and noisy data in nonlinear time-series prediction