In this paper, we consider solving the integer linear systems, i.e., given a matrix A∈Rm×n, a vector b∈Rm, and a positive integer dede2072b1ecf8bf1c6901" title="Click to view the MathML source">d, to compute an integer vector x∈Dn such that Ax≥b or to determine the infeasibility of the system, where m and n denote positive integers, R denotes the set of reals, and D={0,1,…,d−1}. The problem is one of the most fundamental NP-hard problems in computer science.
For the problem, we propose a complexity index η which depends only on the sign pattern of A. For a real edad069bc56ee59835a425" title="Click to view the MathML source">γ, let denote the family of the problem instances I with η(I)=γ. We then show the following trichotomy:
•
is solvable in linear time, if γ<1,
•
is weakly NP-hard and pseudo-polynomially solvable, if γ=1,
•
is strongly NP-hard, if γ>1.
This, for example, includes the previous results that Horn systems and two-variable-per-inequality (TVPI) systems can be solved in pseudo-polynomial time.