Piecewise Linear Value Function Approximation for Factored MDPs.
Pascal PoupartCraig BoutilierRelu PatrascuDale SchuurmansPublished in: AAAI/IAAI (2002)
Keyphrases
- piecewise linear
- factored mdps
- approximate dynamic programming
- dynamic programming
- state space
- basis functions
- markov decision processes
- linear program
- policy iteration
- algebraic decision diagrams
- reinforcement learning
- context specific
- linear combination
- step size
- linear programming
- average cost
- markov chain
- temporal difference
- control policy
- infinite horizon
- hyperplane
- search space
- optimal control
- sufficient conditions
- markov decision problems