Pessimistic Nonlinear Least-Squares Value Iteration for Offline Reinforcement Learning.
Qiwei DiHeyang ZhaoJiafan HeQuanquan GuPublished in: ICLR (2024)
Keyphrases
- reinforcement learning
- nonlinear least squares
- markov decision processes
- state space
- least squares
- optimal policy
- policy iteration
- trust region
- markov decision process
- levenberg marquardt
- image registration
- function approximation
- dynamic programming
- partially observable markov decision processes
- model free
- subpixel accuracy
- neural network
- temporal difference
- optimization methods
- back propagation
- global optimum
- bundle adjustment
- image processing
- learning algorithm