 Now, after all the beautiful theories, let's actually build a classifier for something really real. In this case, we'll take the animal faces data set, which are images. And we, of course, also have the relevant image labels. Now it's time for some real deep learning. Use the generic deep learning function that you built before. And train it on the animal faces data set. And keep in mind that this is not a conf net. See how well it does. Is this surprising?