ConfPad

What all great scientists have in common is the ability to make leaps of insight by way of analogy. In this session, I will talk about the underpinnings of deep learning from a meta-mathematical perspective, and demonstrate how high-level analogies allow us to work more creatively and intelligently. With a set of smouldering GPUs and some nifty code, I will walk you through exactly why patterns such as symmetry, locality, compositionality, and nonlinearity are so easily learned by deep neural networks. Then you too will be able to see the world through the neurons of a deep learning model!