Detecting Spiky Corruption in Markov Decision Processes.
Jason MancusoTomasz KisielewskiDavid LindnerAlok SinghPublished in: AISafety@IJCAI (2019)
Keyphrases
- markov decision processes
- state space
- finite state
- transition matrices
- optimal policy
- planning under uncertainty
- policy iteration
- dynamic programming
- state and action spaces
- decision theoretic planning
- risk sensitive
- finite horizon
- action space
- reinforcement learning
- infinite horizon
- average cost
- reinforcement learning algorithms
- decision processes
- average reward
- reachability analysis
- partially observable
- factored mdps
- markov decision process
- action sets
- model based reinforcement learning
- interval estimation
- reward function
- semi markov decision processes
- stochastic shortest path