Last Thursday, Ben presented two papers on normalizing flows: Rezende and Mohamed, 2015, and Kingma, Salimans, and Welling, 2016. The papers present a scalable way to make the posterior approximation family of variational inference very rich. The key is to put a network of normalizing flows on top of the inference network.

Welcome, and thanks for joining! We’ve got a great lineup of papers on the docket for this semester. Learn more about the reading group here, and come to our first meeting this Thursday, Sept. 22nd, 2-4pm in Stat 1025 to get on the schedule!