Emergence and long-term maintenance of modularity in spiking neural networks with plasticity

In the last three decades the field of brain connectivity has uncovered that cortical regions, interconnected via white-matter fibers, form a modular and hierarchical network. This type of organization, which has also been recognised at the microscopic level in the form of interconnected neural assemblies, is typically believed to support the coexistence of segregation (specialization) and integration (binding) of information. A prominent remaining question is to understand how the brain could possibly become such a complex network. Here, we give a first step into answering this question and propose that adaptation to various inputs could be the key driving mechanism for the formation of structural assemblies at different scales. To illustrate that, we develop a model of (QIF) spiking neurons, subjected to stimuli targetting distributed populations. The model follows several biologically plausible constraints: (i) it contains both excitatory and inhibitory neurons with two classes of plasticity: Hebbian and anti-Hebbian STDP, (ii) dynamics are not frozen after the entrainment is finished but the network is allowed to continue firing spontaneously, and (iii) plasticity remains always active, also after the learning phase. We find that only the combination of Hebbian and anti-Hebbian inhibitory plasticity allows the formation of stable modular organization in the network. Besides, given that the model continues “alive” after the learning, the network settles into an asynchronous irregular firing state displaying spontaneous memory recalls which, as we show, turn crucial for the long-term consolidation of the learned memories.