search

qiufan319 / benchmark_pc_attack

MIT License
11 stars 2 forks source link

How to use the defense framework you presented in your paper？ #1

Open jamesliu609 opened 4 months ago

jamesliu609 commented 4 months ago

Hi, thanks a lot for this work! Where is your defense method and how should I use it? Could you please provide detailed instructions. Also in AT I have a point of confusion, in line 263 the code reads: train_set=torch.utils.data.ConcatDataset([orig_trainset,orig_trainset]). Shouldn't it be org_trainset and adv_trainset that are used for adversarial training, why are two identical org_trainset connected here. Thank you very much if you can answer my confusion!

qiufan319 commented 3 months ago

Hi, thanks a lot for this work! Where is your defense method and how should I use it? Could you please provide detailed instructions. Also in AT I have a point of confusion, in line 263 the code reads: train_set=torch.utils.data.ConcatDataset([orig_trainset,orig_trainset]). Shouldn't it be org_trainset and adv_trainset that are used for adversarial training, why are two identical org_trainset connected here. Thank you very much if you can answer my confusion!

Hi @jamesliu609 , thank you for your correction. The correct code on AT should be: train_set=torch.utils.data.ConcatDataset([orig_trainset,adv_trainset]). And I have updated the hybrid training and instruction in the latest version. If you have any further questions, please feel free to ask me.