From connectome to silicon: a biologically-inspired complex network of CMOS chaotic oscillators for analog brain emulation

The design of networks of physical units coupled via increasingly complex topologies has become a prominent area of interest, partly because of the intricate influences that experimental non-idealities exert on the formation of synchronization patterns. This intersects with the computational study of brain dynamics using large-scale models to recapture the relationship between structural connectivity and the resulting synchronization, encompassed in functional connectivity. This study proposes a new architecture for a CMOS integrated circuit, realizing a dynamical network based on a low-resolution projection of the human structural connectome through instantiating a chaotic oscillator to realize its nodes. A pre-existing compact and area-efficient pure CMOS circuit was selected to realize the node dynamics, the resulting ensemble was simulated using SPICE, and a tentative layout was proposed. The effects of the coupling strength, node-specific parameters, and network topological properties on synchronization and local dynamics were investigated, providing evidence of the non-trivial relationship between structural and functional connectivity. The simulation results of circuit simulations were studied alongside those of a network of Rössler systems, obtaining convergent findings. In particular, the effect of the node degree heterogeneity highlighted the contribution of hub nodes to the collective dynamics. The simulated functional connectivity was projected onto the cortical surface based on the parcel locations used for structural network construction, and an appreciable level of association with the known biological networks was observed. The proposed architecture constitutes a flexible, low-power, minimalistic platform suitable for analog brain simulation, engineering applications such as reservoir computing, and creating in-silico disease models. The complete design materials have been made openly available to support the future physical realization of the integrated circuit.