Policy gradient learning methods for stochastic control with exit time and applications to share repurchase pricing. (arXiv:2302.07320v1 [q-fin.CP]) http://arxiv.org/abs/2302.07320