Global Optimality Guarantees For Policy Gradient Methods.
Jalaj BhandariDaniel RussoPublished in: CoRR (2019)
Keyphrases
- global optimality
- policy gradient methods
- globally optimal
- global optimization
- natural actor critic
- optimal solution
- theoretical guarantees
- discrete optimization
- global minimum
- objective function
- convex functions
- policy gradient
- global solution
- semidefinite
- gradient field
- optimality conditions
- neural network
- worst case
- reinforcement learning