International Encyclopedia of Systems and Cybernetics

G. KLIR states: "This problem is concerned with the question of how a given generative system, assumed to be employed in modeling some aspect of reality, can be broken down into appropriate subsystems. A collection of subsystems, which in our terminology is a structure system, is appropriate when it is sufficient for reconstructing the original overall system to an acceptable degree of approximation".

"Two preference orderings are essential for comparing competing structure systems in the reconstruction problem. One is expressed in terms of the amount of information that is lost when the overall system is replaced by a structure system. Let this preference be called information ordering…

"The second preference ordering involved in the reconstruction problem is connected with the size of subsystems contained in the structure system: smaller subsystems are preferred. This latter preference ordering may be called a complexity ordering since a reduction in the size of a system tends to reduce the size of its description. It is only a partial ordering" (1991, p.106-7).

As to the practical aspect, KLIR states: "Both the reconstruction and identification problem are computationally very hard. As a consequence, the applicability of reconstructability analysis has been limited to systems with a modest number of variables, say a dozen or less. Recent developments of some powerful heuristic procedures by Roger CONANT seem to indicate, however, that systems with hundreds and possibly, even thousands of variables could be analyzed on existing large computers" (p.348).