Complexity of Derivative-Free Policy Optimization for Structured H Control

Neural Information Processing Systems 

Numerical results are also provided to demonstrate our theory.