We would show the network all of the different pictures of cartoon faces, and the pictures of dog faces. The network then goes on to form a generalized picture of all cartoon faces and all dog faces.