Dualizing optimization problems in mathematical economics (Q1060135)

Using the general theory of convex programming, dynamic optimization problems over a countably infinite number of periods are considered. Special forms may show up in particular in mathematical economics. One such form is where the objective function is an infinite series, where the decision variable consists of a countable number of components, such that each term of the series corresponds to a single component in a one- to-one kind of way. Further, there are a countable number of constraints, each containing two successive components of the decision variable. Hence the objective function might be a sum of terms \(f_ t(x_ t)\), and the constraints might be \((x_{t-1},x_ t)\in 0_ t\), \(t=1,2,...\) Models considered before by Weitzman, and by Mitra and Ray can be treated by specializing the general theory of this paper appropriately. Furthermore, the paper considers the choice one has in introducing Lagrange multipliers for all constraints simultaneously, leading to a dual variable of infinitely large dimension (which sometimes could be undesirable), or in introducing them one by one, leading to a sequence of dual problems each with a finite dimensional dual variable. Finally, the paper considers the case where the objective function cannot be defined as indicated above, because the infinite series does not converge, but where the lim inf of the partial sums does exist, and is finite.