The Computational Properties of a Simplified Cortical Column Model

The mammalian neocortex has a repetitious, laminar structure and performs functions integral to higher cognitive processes, including sensory perception, memory, and coordinated motor output. What computations does this circuitry subserve that link these unique structural elements to their function? Potjans and Diesmann (2014) parameterized a four-layer, two cell type (i.e. excitatory and inhibitory) model of a cortical column with homogeneous populations and cell type dependent connection probabilities. We implement a version of their model using a displacement integro-partial differential equation (DiPDE) population density model. This approach, exact in the limit of large homogeneous populations, provides a fast numerical method to solve equations describing the full probability density distribution of neuronal membrane potentials. It lends itself to quickly analyzing the mean response properties of population-scale firing rate dynamics. We use this strategy to examine the input-output relationship of the Potjans and Diesmann cortical column model to understand its computational properties. When inputs are constrained to jointly and equally target excitatory and inhibitory neurons, we find a large linear regime where the effect of a multi-layer input signal can be reduced to a linear combination of component signals. One of these, a simple subtractive operation, can act as an error signal passed between hierarchical processing stages.Author Summary: What computations do existing biophysically-plausible models of cortex perform on their inputs, and how do these computations relate to theories of cortical processing? We begin with a computational model of cortical tissue and seek to understand its input/output transformations. Our approach limits confirmation bias, and differs from a more constructionist approach of starting with a computational theory and then creating a model that can implement its necessary features. We here choose a population-level modeling technique that does not sacrifice accuracy, as it well-approximates the mean firing-rate of a population of leaky integrate-and-fire neurons. We extend this approach to simulate recurrently coupled neural populations, and characterize the computational properties of the Potjans and Diesmann cortical column model. We find that this model is capable of computing linear operations and naturally generates a subtraction operation implicated in theories of predictive coding. Although our quantitative findings are restricted to this particular model, we demonstrate that these conclusions are not highly sensitive to the model parameterization.