Revisiting the Train Loss: an Efficient Performance Estimator for Neural Architecture Search

Open in new window