Fooling Deep Learning Networks

Anh discusses the process of generating images that fool deep learning networks and how these fooling images can generalize across different architectures. The high confidence rates of misclassification raise questions about the fundamental approach to deep learning.