Variational Inference with Normalizing Flows.md

• Variational Inference with Normalizing Flows: https://arxiv.org/abs/1505.05770

One of the best papers I read recently! A very good note can be found here https://casmls.github.io/general/2016/09/25/normalizing-flows.html

An open question, as described in the reference note: ``whether the proposed normalizing flows is a universal approximator of all the distributions when the number of layers go to infinity." Two technical questions: Do we still need to assume P(z_0) as a Gaussian distribution? Why is the model not scalable regarding to high-dimensional latent space?