Moving least-squares approximations for linearly-solvable MDP.
Mingyuan ZhongEmanuel TodorovPublished in: ADPRL (2011)
Keyphrases
- least squares
- policy evaluation
- markov decision processes
- matrix approximation
- policy iteration
- np complete
- parameter estimation
- optimal policy
- robust estimation
- special case
- np hard
- state space
- reinforcement learning
- finite state
- sparse linear
- markov decision process
- levenberg marquardt
- utility function
- planning under uncertainty
- linear program
- markov decision problems
- learning algorithm
- curve fitting
- objective function
- optical flow
- approximation methods
- model free
- optimal solution
- moving objects