Affine Gateaux Differentials and the von Mises Statistical Calculus

Simone Cerreia-Vioglio,Fabio Maccheroni,Massimo Marinacci,Luigi Montrucchio,Lorenzo Stanca
2024-03-13
Abstract:This paper presents a general study of one-dimensional differentiability for functionals on convex domains that are not necessarily open. The local approximation is carried out by affine functionals, rather than linear ones as in standard Gateaux differentiability. This affine notion of differentiability naturally arises in many applications and, here and there, it appeared in the literature. Our systematic analysis aims to give a general perspective on it.
Functional Analysis,Optimization and Control
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the one - dimensional differentiability problem of functional functions defined on convex sets, especially when these convex sets are not necessarily open sets. Specifically, the paper introduces a new concept of differentiability - Affine Gateaux differentiability, which uses affine functionals rather than linear functionals in standard Gateaux differentiability for local approximation. This new concept of differentiability arises naturally in many applications and has important application values in fields such as statistics and economics. The main contribution of the paper lies in systematically analyzing this new concept of differentiability and exploring its applications in optimization, statistics, and economics. For example, the paper proves that weakly affine - differentiable functions are semi - differentiable and gives a mean - value theorem. In addition, the paper also establishes an envelope theorem similar to Danskin, which is applicable to weakly affine - differentiable functions, and discusses specific applications in risk theory and statistics. Through these analyses and results, the paper provides a new perspective and tool for understanding and dealing with functional functions defined on non - open convex sets.