Least Inferable Policies for Markov Decision Processes.
Mustafa O. KarabagMelkior OrnikUfuk TopcuPublished in: ACC (2019)
Keyphrases
- markov decision processes
- optimal policy
- markov decision process
- decision processes
- state space
- reward function
- average cost
- decision problems
- finite horizon
- finite state
- policy iteration algorithm
- discounted reward
- total reward
- decentralized control
- expected reward
- macro actions
- reinforcement learning
- dynamic programming
- policy iteration
- stationary policies
- decision theoretic planning
- average reward
- reachability analysis
- transition matrices
- planning under uncertainty
- multistage
- control policies
- long run
- markov decision problems
- risk sensitive
- partially observable markov decision processes
- reinforcement learning algorithms
- partially observable
- model based reinforcement learning
- initial state
- semi markov decision processes
- infinite horizon
- stochastic shortest path
- action space
- action sets
- state and action spaces
- factored mdps
- sufficient conditions
- search space
- objective function