Aggregation in large-scale optimization (Q1412415): Difference between revisions

The book is devoted to aggregation and disaggregation techniques, which are used to manage large-scale optimization problems. The material is divided into 3 chapters. If we replace the solution of an original large-scale problem by the solution of an aggregated problem having a smaller size, it is necessary to estimate the error, which arises from this replacement without solving the original large problem. In Chapter 1, two types of the bounds of the aggregation error are considered -- a priori and a posteriori. While the a priori bounds are calculated before solving the aggregated problem, the a posteriori bounds are derived from the optimal solution of the aggregated problem. The aggregation error bounds are calculated for linear, convex, and special-structured problems as well as for some integer programming problems. The aggregation depends on certain aggregation parameters. Each fixed values of these parameters correspond to an aggregation problem connected with an aggregation error depending on the chosen values of aggregation parameters. Chapter 2 is devoted to the problem of iterative adjustment of the values of the aggregation parameters, which reduces the aggregation error. This procedure is called iterative aggregation. Relations of this procedure to the classical decomposition schemes are investigated. Chapter 3 deals with the aggregation in parametric optimization problems. In these problems, the optimal value of the objective function depends on the parameters. The problem of the identification of parameter values resulting in a better optimal value of the objective function is studied. The authors investigate further so called consistent aggregation with respect to parameter variation. This property of an aggregation can be characterized as follows: if a change of parameters results in an improvement of the objective function value of the aggregated problem, then this change improves also the value of the objective function in the original problem and vice versa. The consistent aggregation is investigated for linear and convex programming problems, as well as for some optimal control and integer programming problems. Theoretical explanations are accompanied with illustrative examples and comments and references at the end of each chapter.