Policy Learning of MDPs with Mixed Continuous/Discrete Variables: A Case Study on Model-Free Control of Markovian Jump Systems. (arXiv:2006.03116v1 [math.OC]) http://arxiv.org/abs/2006.03116