Convergence of the Deep Galerkin Method for Mean Field Control Problems