Data-Driven Control of Nonlinear Systems: Learning Koopman Operators for Policy Gradient | IEEE Conference Publication | IEEE Xplore