Learning algorithms for Markov decision processes with average cost

SIAM Journal on Control and Optimization

Volumn 40, Issue 3, 2002, Pages 681-698

  Abounadi, Jinane ^a   Bertsekas, Dimitrib ^a   Borkar, V S ^a  

^a MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

Author keywords

Average cost control; Controlled Markov chains; Dynamic programming; Q learning; Simulation based algorithms; Stochastic approximation

Indexed keywords

COMPUTER SIMULATION; COSTS; DECISION THEORY; DYNAMIC PROGRAMMING; LEARNING ALGORITHMS; OPTIMAL CONTROL SYSTEMS;

AVERAGE COST CONTROL;

MARKOV PROCESSES;

EID: 0036287773     PISSN: 03630129     EISSN: None     Source Type: Journal    
DOI: 10.1137/S0363012999361974     Document Type: Article

References (29)

1. Distributed dynamic programming
2. A new value iteration method for the average cost dynamic programming problem
3. Rollout algorithms for stochastic scheduling problems
4. Stochastic approximation with two time scales
5. Recursive self-tuning control of finite Markov chains
6. Asynchronous stochastic approximations
7. A learning algorithm for discrete time stochastic control
8. On the number of samples required for Q-learning
9. The ODE method for convergence of stochastic approximation and reinforcement learning
10. A new analog parallel scheme for fixed point computation, part I: Theory
11. On the convergence of stochastic iterative dynamic programming algorithms
12. Adaptive control of Markov chains with local updates
13. Actor-critic-type learning algorithms for Markov decision processes
14. Average reward reinforcement learning: Foundations, algorithms and empirical results
15. A reinforcement learning method for maximizing undiscounted rewards
16. Reinforcement learning algorithms for average payoff Markovian decision processes
17. Asynchronous stochastic approximation and Q-learning
18. Feature-based methods for large scale dynamic programming
19. Q-learning
20. Smoothing derivatives of functions and applications