As a complete layman reading that paper, Skilling's arguments seem quite damning of IG. In particular, Amari's fundamental assumption seems to be refuted: "For

As a complete layman reading that paper, Skilling's arguments seem quite damning of IG. In particular, Amari's fundamental assumption seems to be refuted: "For inference, the only acceptable value for the Rényi-Tsallis parameter is α = 1, which is the correct information. That negates the generalisation to α != 1 which underlies Amari’s “α-divergences” in information geometry." Have any IG proponents responded to or refuted Skilling's critiques? This is interesting because Shun'ichi Amari is credited amongst others with advancing the field in the 80's[1]. [1] https://en.m.wikipedia.org/wiki/Information_geometry https://en.m.wikipedia.org/wiki/Information_geometry