Circuit Models in Neuroblox

Introduction

We saw how Bloxs and their connection rules are defined and we can connect external sources to single neuron or neural mass Bloxs. Now we are ready to expand our models to resemble more closely the neuronal circuits of the brain.

This session includes two examples of circuit models

• a biomimetic corticostriatal model of visual processing

• a Pyramidal-Interneuron Gamma Network (PING) of Excitation-Inhibition balance.

Learning goals :

• build complex models from the systems neuroscience literature.

• introduce composite Bloxs with hierarchical structures.

• use more advanced plotting recipes to visualize complex Bloxs and circuits of them.

Namespaces

When defining our own Bloxs previously or when creating Blox objects we left the namespace=nothing keyword argument to its default value. This is because it was unnecessary to provide an explicit namespace qualifier, as we were working with "atomic" Bloxs, that is Bloxs that contained some dynamics of a neuron or a neural mass model.

Neuroblox also contains composite Bloxs though, as subtypes of CompositeBlox. All CompositeBlox subtypes do not have their own dynamics, but they contain either other CompositeBloxs or Bloxs with dynamics such as Neuron and/or NeuralMass types. When working with CompositeBloxs with multiple layers its important to declare a namespace value, which will be the outermost name of our model. Neuroblox uses this namespace to accumulate all connection terms and match them correctly to the appropriate input variable of each Blox.

Example

Let's consider two WinnerTakesAllBlox objects which are <: CompositeBlox and include multiple excitatory neurons and a feedback inhibitory neuron implementing lateral inhibition.

model_name = :g
@named wta1 = WinnerTakeAllBlox(namespace = model_name)
@named wta2 = WinnerTakeAllBlox(namespace = model_name)

then wta1₊exci1₊V and wta2₊exci2₊V are the voltage variables (membrane potentials) of the first excitatory neuron (HHNeuronExciBlox) in wta1 and wta2 respectively. Each ₊ character adds another namespace if read from right to left.

We have chosen model_name = :g as the name of our model that contains these two WTA Bloxs. This name will be given to the name field of system_from_graph that creates the final model with all connection equations and simplified structure. We provide an explicit name instead of using the @named macro.

g = MetaDiGraph()
add_edge!(g, wta1 => wta2, weight=10)

sys = system_from_graph(g, name = model_name)

so after we turn the graph into a system, the above variables will be g₊wta1₊exci1₊V and g₊wta2₊exci2₊V.