Reusing Historical Trajectories in Natural Policy Gradient via Importance Sampling: Convergence and Convergence Rate.
Yifan LinYuhao WangEnlu ZhouPublished in: CoRR (2024)
Keyphrases
- convergence rate
- importance sampling
- policy gradient
- gradient method
- variance reduction
- monte carlo
- convergence speed
- step size
- kalman filter
- particle filter
- learning rate
- markov chain
- particle filtering
- approximate inference
- function approximation
- reinforcement learning
- cost function
- machine learning
- posterior distribution
- markov chain monte carlo
- visual tracking
- moving objects