Primal-dual regression approach for Markov decision processes with general state and action space.
Denis BelomestnyJohn SchoenmakersPublished in: CoRR (2022)
Keyphrases
- action space
- markov decision processes
- primal dual
- state space
- reinforcement learning
- state and action spaces
- finite state
- optimal policy
- state action
- linear programming
- continuous state
- convex optimization
- special case
- real valued
- dynamic programming
- average reward
- linear program
- approximation algorithms
- markov decision process
- policy iteration
- partially observable
- average cost
- reinforcement learning algorithms
- markov chain
- single agent
- stochastic processes
- state variables
- convergence rate
- search space
- infinite horizon
- decision makers
- function approximators
- markov decision problems
- dynamical systems
- continuous state spaces