Answers are grounded in this paper and cite their sources.
This paper studies the inverse problem in information geometry: given a Hessian manifold, is it derived from an open exponential family? As a motivating case, we prove that for the natural exponential family generated by the generalized hyperbolic secant distribution (NEF-GHS) with shape parameter $$p>0$$ p > 0 , the dual Hessian manifold is not derived from any open exponential family, thereby extending a result of Letac for the case $$p=1$$ p = 1 . Our proof exploits a local obstruction coming from branch points in the Hessian potential. Inspired by this observation, we prove a broad result: when a Hessian metric admits a real-analytic global Hessian potential, “derivedness” is determined by its restrictions to open submanifolds.
No takes yet. Share an insight, caveat, or question.
Keiji Yahata (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: