Global Convergence of Policy Gradient Methods to (Almost) Locally Optimal Policies.
Kaiqing ZhangAlec KoppelHao ZhuTamer BasarPublished in: CoRR (2019)
Keyphrases
- global convergence
- optimal policy
- policy gradient methods
- reinforcement learning
- natural actor critic
- policy gradient
- global optimum
- convergence rate
- optimization methods
- convergence speed
- convergence analysis
- average reward
- markov decision processes
- decision problems
- dynamic programming
- actor critic
- state space
- partially observable markov decision processes
- infinite horizon
- policy iteration
- finite state
- sufficient conditions
- long run
- function approximation
- average cost
- particle swarm
- hybrid algorithm
- initial state
- optimal control
- reinforcement learning methods
- optimization method
- objective function
- machine learning