There is an interesting similarity between the way that neural networks learn about data and the way that models of particle production are structured.  This connection between neural networks and renormalisable theories may help explain why deep learning has been successful for a wide range of tasks.  In this talk I will describe a neural network that has been structured so as to behave as a toy parton shower model for particle production, which may make some of the connections between deep learning and particle physics more explicit.