A Primal-Dual Algorithm for Offline Constrained Reinforcement Learning with Low-Rank MDPs.
Kihyuk HongAmbuj TewariPublished in: CoRR (2024)
Keyphrases
- low rank
- reinforcement learning
- markov decision processes
- convex optimization
- linear combination
- missing data
- matrix factorization
- state space
- low rank matrix
- matrix completion
- optimal policy
- rank minimization
- singular value decomposition
- function approximation
- high dimensional data
- semi supervised
- matrix decomposition
- reinforcement learning algorithms
- markov decision process
- minimization problems
- trace norm
- high order
- model free
- kernel matrix
- action space
- low rank matrices
- robust principal component analysis
- learning problems
- singular values
- policy iteration
- reward function
- temporal difference
- dynamic programming
- learning algorithm
- nuclear norm
- learning process
- function approximators
- markov decision problems
- pattern recognition
- neural network
- data sets