基于pytorch的Fast Gradient Sign Attack (FGSM)在mnist数据集上的伪造(原理近似gan,但是是求loss在图像上的梯度来改变图像))
%matplotlibinlineAdversarialExampleGenerationAuthor:NathanInkawhich__Ifyouarereadingthis,hopefullyyoucanappreciatehoweffectivesomemachinelearningmodelsare.ResearchisconstantlypushingMLmodelstobefaster