Pessimistic Nonlinear Least-Squares Value Iteration for Offline Reinforcement Learning.
Qiwei DiHeyang ZhaoJiafan HeQuanquan GuPublished in: CoRR (2023)
Keyphrases
- reinforcement learning
- nonlinear least squares
- markov decision processes
- least squares
- state space
- policy iteration
- trust region
- optimal policy
- markov decision process
- levenberg marquardt
- image registration
- dynamic programming
- function approximation
- partially observable markov decision processes
- subpixel accuracy
- global optimum
- model free
- machine learning
- optimization methods
- optical flow
- learning algorithm
- temporal difference
- fixed point
- maximum likelihood