Metastability/Criticality as an evolutionary set-point for cortical dynamics

Research in the past 20 years has increasingly suggested that cortical dynamics are tuned to a regime where several computational functions are optimized, often formalized as critically, following frameworks from statistical physics and the study of chaos. More importantly, it appears that cortical network attain this state through processes of self-organization. One such processes, shaping criticality at both the circuit and large-scale levels, is the homeostasis of excitatory-inhibitory balance, through the regulation of firing rates in pyramidal neurons. Average firing rates are remarkably stable across time and even cortical areas, suggesting they are fine tuned at the neuronal level to maintain network criticality. However, it is unclear how this fine-tuning occurs within single organisms, given that individual neurons do not have direct access to information on the dynamics of the whole cortex.

To solve this conundrum, we propose that, since cortical criticality has computational benefits, it shapes behavior in a way that improves organism fitness. Therefore, we hypothesize that local circuitry and average firing rates of the cortex should be tuned through evolution in a way that brings cortical dynamics closer to criticality.

To test this hypothesis, we will use large-scale neural mass model of the cortex with biophysically-inspired local dynamics. Then, taking local parameters such as mean firing rates and the timescales of excitatory and inhibitory neuronal populations as a proxy for the "genetic code" of each model, we will evolve naive models using an evolutionary algorithm that takes large-scale criticality as a fitness function and accounts for constraints such as the minimization of energy consumption.

With this effort, we aim to demonstrate that, by assuming criticality as an evolutionary set-point for cortical dynamics, our models will evolve local circuits in line with the canonical microcircuits of the mammalian cortex, demonstrating that these processes of self-organization extend beyond the lifetime of a single individual into evolutionary timescales.