The network programming method in control over target programs

Abstract

We consider the problem of optimizing programs with respect to cost. The state of a program is defined via several goals (criteria) based on a complex evaluation system. For each goal, there exist several projects that contribute to achieving this goal. A number of projects contributes to several goals (multipurpose projects). To solve the problem, we propose two methods. One is based on enumerating all possible ways to include multipurpose projects into the program. The other is based on a network programming technique. We give results of numerical experiments that compare these approaches.