Opening the Black Box: predicting the trainability of deep neural networks with reconstruction entropy

Open in new window