Quantum states can be highly complex objects to model classically. What are the boundaries of neural networks representing these states? How well can they approximate? How can complex states be efficiently approximated and how do they connect to the complexity zoo?