An Alternate Policy Gradient Estimator for Softmax Policies.
Shivam GargSamuele TosattoYangchen PanMartha WhiteRupam MahmoodPublished in: AISTATS (2022)
Keyphrases
- policy gradient
- policy gradient methods
- policy search
- variance reduction
- partially observable markov decision processes
- reinforcement learning
- actor critic
- function approximation
- optimal control
- natural actor critic
- gradient method
- least squares
- temporal difference learning
- model free reinforcement learning
- optimal policy
- approximation methods
- average reward
- monte carlo
- maximum likelihood
- reinforcement learning algorithms
- importance sampling
- dynamical systems
- reinforcement learning methods
- finite state
- naive bayes classifier