3 ms·
The Darker Side of Machine Learning
- jszymborski 9y agoQuestion for HN people who know better: how generalisable are these attacks on deep learning networks. I know it's trivial to optimise for a class on say the InceptionV3 networks trained on ImageNet; but can I perform the same kind of attack if I don't necessarily know the what the network looks like? What if I don't know the details of the dataset it was trained on?