Gradient descent follows the regularization path for general losses

Open in new window