search

yaodongyu / TRADES

TRADES (TRadeoff-inspired Adversarial DEfense via Surrogate-loss minimization)
MIT License
510 stars 123 forks source link

trades.py #6

Closed liuye6666 closed 5 years ago

liuye6666 commented 5 years ago

I have read your article many times and feel that the article is well written, but I have a question. Should the sample be standardized to [0, 1]? Can it be standardized to [-1, 1]?why? Thank you very much,。

hongyanz commented 5 years ago

Thanks for your interests in our paper. There is no requirement that the sample should be standardized to [0, 1] and it can be standardized to other ranges as well, although in our codes/experiments we choose the former one as the PyTorch standardizes samples to [0, 1] by default.