This was part of Advances in Optimal Decision Making under Uncertainty

Analysis and implementation of policy gradient methods for continuous-time stochastic control

Christoph Reisinger, University of Oxford