Noise-induced maximization of integrated information in FitzHugh-Nagumo networks

Increasing evidence suggests that consciousness is linked to a system's ability to integrate information. Integrated information measures how much information a system holds beyond the sum of its independent parts. In this study, we explore how noise can maximize integrated information in small-scale FitzHugh-Nagumo neural networks. This noise-driven optimization occurs in both oscillatory and excitable regimes, as well as at the bifurcation point. Furthermore, we observe a noise-induced correlation between integrated information and coherence, and notably, the maximization of integrated information is accompanied by disparate coherence patterns in different regimes. Our findings provide insights into how noise can enhance information processing and potentially consciousness in neural systems and reveal a connection between information integration and the regularity of neural activity.