Feature Selection Using Regularization in Approximate Linear Programs for Markov Decision Processes.
Marek PetrikGavin TaylorRonald ParrShlomo ZilbersteinPublished in: ICML (2010)
Keyphrases
- markov decision processes
- linear program
- factored mdps
- dynamic programming
- linear programming
- policy evaluation
- policy iteration
- state space
- finite state
- optimal policy
- optimal solution
- average cost
- simplex method
- column generation
- stochastic programming
- reinforcement learning algorithms
- objective function
- reinforcement learning
- decision theoretic planning
- planning under uncertainty
- average reward
- transition matrices
- primal dual
- support vector
- partially observable
- infinite horizon
- machine learning
- markov decision process
- action space
- decision processes
- reward function
- np hard
- action sets
- partially observable markov decision processes
- stochastic games
- markov decision problems
- stationary policies