We discussed MAF (arXiv:1705.07057v4) last time: The paper did not explain how exactly is MADE being used to update the shift and logscale.
We will use the tensorflow implementation of MAF to probe the above question. Here is the link to the relevant documentation: https://www.tensorflow.org/probability/api_docs/python/tfp/bijectors/MaskedAutoregressiveFlow
Refer to references.
Table of Contents