Compressive sensing with un-trained neural networks: Gradient descent finds the smoothest approximation

Open in new window