dongyp13/adversarial-distributional-training

Adversarial Distributional Training (NeurIPS 2020)

adversarial-distributional-training's Issues

Could you release the code for the visualization part (Figure. 1 and 2) in your paper?

Can 2080Ti GPU support the computation power?

Just for confirm, I try to run default adt_expam algorithm on a leasehold server, however, the connection is closed. I want to know whether it is the problem of server itself or whether the single-GPU of 2080Ti cannot support the batch-size of GPU memory on the default setting.

Moreover, thanks for you sharing and contribution.

Regards!
Momo

Calculation of per-example accuracy

First of all, I would like to thank you for this incredible work!

Seems that you only present the robustness evaluation per attack in the testing. May I know how do you calcualte the per-example accuracy (A_rob in the paper)?

Thanks.