Hello, thanks for this brilliant work.
In your paper, more specifically, Section 4.1 CIFAR, you mentioned: "for the final run we use all 50000 images and report the final test error at the end of training". What does this mean? The training was repeated several times, and each time with the weights from the previous run? And in the last run, training Densenet without validation, and getting the final result on the 10000 testing images?
Hello, thanks for this brilliant work. In your paper, more specifically, Section 4.1 CIFAR, you mentioned: "for the final run we use all 50000 images and report the final test error at the end of training". What does this mean? The training was repeated several times, and each time with the weights from the previous run? And in the last run, training Densenet without validation, and getting the final result on the 10000 testing images?
Best wishes, Qiang