Lecture 3 : Intro to Deep Learning

In the classification task, given some data x(i), we want to produce a class label. Suppose there are K possible classes for the data points. In this setting, suppose that the labels are encoded as one-hot, meaning each y(i) is a vector of length K with a 1 in the entry corresponding to the correct class and 0 in every other entry. Then, we can consider y(i) to be a distribution over the class labels.