search

MadryLab / cifar10_challenge

A challenge to explore adversarial robustness of neural networks on CIFAR10.
MIT License
488 stars 133 forks source link

Same accuracy logged twice #14

Closed JonathanCMitchell closed 5 years ago

JonathanCMitchell commented 5 years ago

Why is the accuracy logged out twice to tensorboard for the adversarial samples in https://github.com/MadryLab/cifar10_challenge/blob/master/train.py#L76. It seems that these are just duplicates.

dtsip commented 5 years ago

Yes, they are duplicates, but we log them under different tags. This is useful when training a bunch of models and you want to compare all the training accuracies in one plot but also be able to compare train and test accuracy of a single model in one plot.