damianliumin / non-adversarial_backdoor Goto Github PK

Dear authors,

I am interested in your method "Non-adversarial Backdoor” and I would like to reproduce your experiments.

However, I found that your code lacks some critical details that are necessary for understanding and replicating your method. Specifically, you do not describe the three backdoor detection techniques and the two relabeling strategies that you use in your experiments. Without them, it is difficult to compare your method to other methods.

Could you please provide more information about these techniques and strategies in your paper or in this repo? This would help me and other researchers to better appreciate your contribution and assess its validity.

Thank you for your attention and cooperation. I look forward to hearing from you soon.