Agnostic Reinforcement Learning with Low-Rank MDPs and Rich Observations.
Ayush SekhariChristoph DannMehryar MohriYishay MansourKarthik SridharanPublished in: NeurIPS (2021)
Keyphrases
- low rank
- reinforcement learning
- markov decision processes
- missing data
- convex optimization
- linear combination
- matrix factorization
- state space
- low rank matrix
- optimal policy
- singular value decomposition
- rank minimization
- matrix completion
- high order
- function approximation
- reinforcement learning algorithms
- semi supervised
- high dimensional data
- kernel matrix
- partially observable
- markov decision process
- low rank matrices
- matrix decomposition
- non rigid structure from motion
- robust principal component analysis
- model free
- learning algorithm
- reward function
- policy iteration
- minimization problems
- machine learning
- higher order
- trace norm
- nearest neighbor
- dynamic programming
- markov decision problems
- neural network