An Exhaustive Study of Two-Node McCulloch-Pitts Networks

Autor: Li, Wentian, Provata, Astero, MacCarthy, Thomas
Rok vydání: 2024
Předmět:
Druh dokumentu: Working Paper
Popis: Boolean networks are widely used in computational biology, evolutionary studies, and social sciences. However, the set of all Boolean-function-defined networks are harder to study as a whole. On the other hand, McCulloch-Pitts gates are sparsely parameterized using only a few number of link strengths, making it possible to study and compare different networks models. We treat two-node McCulloch-Pitts systems as a minimal complex system. When the link strengths are discretized, $3^4=81$ network models or rules are organized in the rule space The limiting dynamics of each rule may depend on the choice of binary state value ([-1,1] or [0,1]), and on the treatment at the threshold point, leading to at least six variants. One variant with [-1,1] as the binary state value (V1 model) tends to have a more diverse dynamical behaviors with a mixture of multiple cycles and fixed points at the limiting state, whereas other variants tend to fall only to fixed-point limiting dynamics. We use V1 models to study the organization of rules with different dynamics in the rule space and robustness of limiting dynamics with respect to a mutation in the rule table, as well as the related phenomena of phase transition and edge-of-chaos. We use another variant (V4 models) with only the fixed-point limiting dynamics to study the robustness of limiting state with respect to perturbation of initial states. The two types of robustness do not seem to be associated with each other. Other aspects of fully discretized two-node MaCulloch-Pitts networks are also studied, including: the proposal of a seventh variant based on a difference equation; relation to Rene Thomas' two types of feedback loops; spectrum properties of state space transition matrix; and asynchronous updating. Our works also expand the concept of network motifs by allowing more finer details.
Comment: 5 figures
Databáze: arXiv