LinkedIn interview question

What is batch normalization? What is the loss function for an SVM? Why use SVM over neural nets?