CROP: Certifying Robust Policies for Reinforcement Learning through Functional Smoothing.
Fan WuLinyi LiZijian HuangYevgeniy VorobeychikDing ZhaoBo LiPublished in: ICLR (2022)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- machine learning
- computationally efficient
- markov decision processes
- robust statistical
- neural network
- control policies
- markov decision process
- state space
- information retrieval
- partially observable markov decision processes
- markov decision problems
- robust statistics
- smoothing methods
- multi agent reinforcement learning
- dynamic programming