Information thermodynamics in stochastic Boolean networks
Complex systems and Biological physics seminar
Tuesday 17 September 2019
Shun Otsubo (University of Tokyo)
Complex networks have attracted much attention in various fields, and a variety of analysis methods have been adopted. One approach is to focus on small subgraph patterns that can be seen as building blocks of a large-scale network. While many studies have investigated the functional roles of such patterns, their thermodynamic aspects have not been much explored. In a different context, information thermodynamics has been developed to quantify dissipation of fluctuating subsystems where information flow plays a crucial role. Here, we consider a simple model called stochastic Boolean network, and propose a systematic characterization of subgraph patterns on the basis of information thermodynamics. Specifically, we focus on three-node patterns, which receive one or two input signals that carry external information. For the case of a single input, we found that all the three-node patterns are classified into four types by using information-thermodynamic measures such as dissipation and mutual information. Next, we consider the case that there are two input, and evaluate the capacity of logical operation of the three-node patterns by using tripartite mutual information, and found that patterns with fewer edges make logical operation efficient.