Fig. 2.

Canonical microcircuit (CMC). a) A patch of cortex divided into cortical columns. Electrical activity of each cortical column can be captured by electrophysiological recording, e.g., ECoG. Each cortical column is divided into several layers (here 3), each of which is modelled by one population of neurons. Superficial and deep pyramidal cells are in the superficial and deep layers, respectively. Excitatory interneurons (spiny stellate cells) are located in layer four – labelled 1 in the figure. Inhibitory interneurons are distributed across all layers and are modelled using one population that interacts with all other populations. b) The mean electrical activity of each neuronal population is derived using mean field theory using two conversion operators. The postsynaptic potentials, ${\mathit{x}}_{\mathit{i}}$ are transformed through a sigmoid nonlinearity, $\mathit{s}(.)$, to generate a firing rate (weighted by connectivity constant g). The ensuing firing rate is then converted to postsynaptic potentials, ${\mathit{x}}_{\mathit{o}}$ through the linear response of synapses (parametrised by rate constant T). In addition, each population is equipped with a self-inhibition connection G (illustrated as short curved red lines) which assures neuronal homeostasis, i.e., in the absence of neuronal input, the activity of neurons rest at an equilibrium. Please see Appendix D for detailed formulation of this model.