Existence, uniqueness, and convergence rates for gradient flows in the training of artificial neural networks with ReLU activation | Publicación