ReSS: A tool for discovering relevant sets in complex systems

A complex system can be composed of inherent dynamical structures, i.e., relevant subsets of variables interacting tightly with one another and loosely with other subsets. In the literature, some effective methods to identify such relevant sets rely on the so-called Relevance Indexes (RIs), measuring subset relevance based on information theory principles. In this paper, we present ReSS, a collection of CUDA-based programs computing two of such RIs, either through an exhaustive search or a niching metaheuristic when the system dimension is too large. ReSS also includes a script that iteratively activates the search and identifies hierarchical relationships among the relevant subsets. The main purpose of ReSS is to establish a common and easy-to-use general RI-based platform for the analysis of complex systems and other possible applications.