DeepLearning3.3, Why are there so many saddle points?

From Wulfram Gerstner  

views comments