Q-learning for risk-sensitive control

Mathematics of Operations Research

Volumn 27, Issue 2, 2002, Pages 294-311

  Borkar, V S ^a  

^a TATA INSTITUTE OF FUNDAMENTAL RESEARCH   (India)

Author keywords

Dynamic programming; Markov decision processes; Q learning; Reinforcement learning; Risk sensitive control; Stochastic approximation

Indexed keywords

BOUNDARY CONDITIONS; COMPUTER SIMULATION; CONVERGENCE OF NUMERICAL METHODS; DECISION THEORY; DYNAMIC PROGRAMMING; LEARNING ALGORITHMS; LEARNING SYSTEMS; MATRIX ALGEBRA; ORDINARY DIFFERENTIAL EQUATIONS; RISK ASSESSMENT; THEOREM PROVING;

MARKOV DECISION PROCESSES; Q-LEARNING ALGORITHM; REINFORCEMENT LEARNING; RISK SENSITIVE CONTROL; STOCHASTIC APPROXIMATION ALGORITHM;

MARKOV PROCESSES;

EID: 0036577013     PISSN: 0364765X     EISSN: None     Source Type: Journal    
DOI: 10.1287/moor.27.2.294.324     Document Type: Article

References (35)

1. Stochastic approximation for nonexpansive maps: Application to Q-learning algorithms
2. Learning algorithms for Markov decision processes with average cost
3. Multiplicative ergodicity and large deviations for an irreducible Markov chain
4. A learning algorithm for Markov decision processes with adaptive state aggregation
5. Neuron-like elements that can solve difficult learning control problems
6. Risk sensitive dynamic asset management
7. Risk sensitive asset management with transaction costs
8. Risk sensitive control of finite state Markov chains in discrete time, with applications to portfolio management
9. Stochastic approximation with two time scales
10. Asynchronous stochastic approximation
11. A learning algorithm for discrete time stochastic control
12. Risk sensitive optimal control for Markov processes with monotone cost
13. The O.D.E. method for convergence of stochastic approximation and reinforcement learning
14. Connections between stochastic control and dynamic games
15. Risk-sensitive control of discrete time Markov processes with infinite horizon
16. Risk-sensitive control of finite state machines on an infinite horizon I
17. Risk sensitive control of Markov processes in countable state space
18. Actor-critic type learning algorithms for Markov decision processes
19. On the convergence of stochastic iterative dynamic programming algorithms
20. Analysis of recursive stochastic algorithms
21. Between MDPs and semi-MDPs: Learning, planning and representing knowledge as multiple temporal scales
22. Asynchronous stochastic approximation and Q-learning
23. Feature-based methods for large scale dynamic programming
24. An analysis of temporal difference learning with function approximation
25. Learning and value function approximation in complex decision processes
26. Neuro-dynamic programming: Overview and recent trends
27. Learning from delayed rewards
28. Q-learning