Deep neural networks algorithms for stochastic control problems on fi nite horizon Les Cahiers de la Chaire / N°78