Damage in simulated neural networks: impact of neuronal aggregation

Abstract

Here, we numerically modelled biologically-realistic neuronal networks. We considered neurons that connected to one another on a Euclidean space and used the Izhikevich model to describe their the activity. Inhibitory and excitatory neurons were considered, and were positioned on the Euclidean space in either a homogeneous or aggregated way. Axons emerging from them were modelled as random walkers. Once the network was built, targeted and random damage were applied, and the dynamic response of the network was quantified, measuring the impact of damage using network analysis. Results show that the simulated networks are most resilient when random attack is applied and nodes are arranged on an aggregated way. The change in dynamics exhibits a non-trivial behaviour, as it is heavily dependent not only on the type of damage applied, but also on the way the network is created and the type of neurons that are deleted.