Supported Trust Region Optimization for Offline Reinforcement Learning.
Yixiu MaoHongchang ZhangChen ChenYi XuXiangyang JiPublished in: ICML (2023)
Keyphrases
- trust region
- reinforcement learning
- optimization methods
- unconstrained optimization
- line search
- optimization problems
- column generation
- optimization algorithm
- optimization method
- constrained optimization
- global optimum
- hessian matrix
- function approximation
- convergence rate
- global convergence
- newton method
- state space
- machine learning
- learning algorithm
- risk minimization
- multi objective
- log likelihood
- quadratic programming
- objective function
- loss function
- simulated annealing
- dynamic programming