We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
总结：对抗样本的开山之作，发现向目标样本添加难以察觉的扰动可以显著影响神经网络的结果。