(Convex) level sets integration (Q504802)

The authors investigate the problem of recovering a pseudoconvex function from a multi-valued map that generates the normal cones to the (lower) level set of the function. This ``(convex) level set integration'' is a generalization of the classical problem of recovering (integrating) a differentiable function from its gradient with a given value in some point. Pseudoconvexity of the seeking function is a generalization of convexity but narrower than quasiconvexity, and multiple-valuedness of the mentioned map is the generalization of the function's differentiability to the subdifferentiability. Such generalized properties occur in optimization and modeling of economic events. The authors point on the revealed preference problem as an important application (changing pseudoconvexity on pseudoconcavity). They understand it as recovering a utility function rationalizing the known demand function. We note that the demand defined completely on a region of interest in the goods' space is the main object in economics, and a utility function rationalizing the demand is only a mathematical tool providing the demand calculations in every point of the region. In applications, the problem of recovering (construction) a utility function arises when only finite values of the demand (trade statistics) are known. In this practical case, according to the \textit{H. R. Varian}'s variant [Econometrica 50, 945--973 (1982; Zbl 0483.90006)] of the famous \textit{S. N. Afriat}'s theorem [Int. Econ. Rev. 8, 67--77 (1967; Zbl 0248.90001)], the existence of a nonsatiated utility function that rationalizes a finite number of the demand data is equivalent to the existence of a nonsatiated, continuous, concave, monotonic utility function that rationalizes this data. However, pseudoconvexity or pseudoconcavity are attractive properties of functions under optimization, and their non-differentiability is a frequent property in applications. Thus, the connection of such functions with their subdifferentials is interesting for generalized convex analysis and constrained optimization. The general convex level sets' integration problem can be connected with minimization of a pseudoconvex function on a convex set. This problem consists of finding a \(g\)-pseudoconvex function \(f\) (the notion `\(g\)-pseudoconvexity' was introduced by \textit{J.-P. Crouzeix} et al. in [Math. Program. 123, No. 1 (B), 61--83 (2010; Zbl 1189.26019)]) such that all values of a given cyclically pseudomonotone multivalued map belong to values of the normal cone map associated with \(f\). This general problem is deeply investigated in the paper concerning existence and uniqueness. The obtained results are applied to the problem of recovering a utility function rationalizing the known demand function.