In a recent work arXiv:1910.11163, we investigated the learning dynamics of neural quantum states under stochastic reconfiguration (quantum version of the natural gradient) using information geometry (measured with quantum Fisher information).

For interested people, I uploaded a presentation file from my talk in the Group seminar here. It contains an introductory comparision between generative models in machine learning and variational quantum Monte-Carlo for readers who are not familiar with this topic, and some movies that show the learning dynamics more clearly.