Abstract: | In this paper we present a way to interpret column aggregation schemes in linear programming as a special kind of primal decomposition. This relation between aggregation and decomposition is obtained through a reformulation of the original problem by the introduction of auxiliary variables. The relation between aggregation and decomposition yields a natural iterative aggregation scheme, where weights updating can be done in different ways. We describe several weight updating schemes and illustrate three of them within an iterative aggregation technique with a numerical example. Finally we point out some new research issues that appear when the aggregation process is viewed in this decomposition framework |