Deep Primal-Dual Reinforcement Learning: Accelerating Actor-Critic using Bellman Duality.
Woon Sang ChoMengdi WangPublished in: CoRR (2017)
Keyphrases
- actor critic
- primal dual
- linear programming
- linear program
- approximate dynamic programming
- reinforcement learning
- duality gap
- gradient method
- convergence rate
- temporal difference
- policy iteration
- optimal control
- policy gradient
- interior point methods
- reinforcement learning algorithms
- dynamic programming
- semidefinite programming
- neuro fuzzy
- algorithm for linear programming
- convex optimization
- optimal solution
- approximation algorithms
- step size
- optimal policy
- average cost
- np hard
- markov decision processes
- transfer learning
- objective function
- multi agent
- function approximation
- average reward
- function approximators
- supervised learning
- markov decision process
- reward function
- model free