Sequential composition of clans in linear systems

Abstract

To obtain an additional computational speedup in the course of solution of linear systems, it is proposed to organize a sequential process of composition of their clans. Computational speedup was obtained through the solution of a sequence of clan composition systems with essentially lower dimensions using the graph of system decomposition into clans. A comparative analysis of sequential composition for subgraphs and edge (paired 3) composition was performed. The problem of construction of systems sequence with the lowest dimension was named by a collapse of weighted graph. The upper and lower limits of the collapse width which corresponds to the dimension of systems were estimated. A heuristic algorithm of optimal collapse was constructed and statistically grounded.