Another big step at this part of the story was moving from a more complicated proof based on functional derivatives to the simpler one based on convexity of KL divergences that we published. @dwf contributed this final step
Simplified KL Divergence Proof Advances Machine Learning Theory
By
–