Populations

After this page you can build a population, choose its neuron, synapse and spike models, and know where connections deliver their input. The equations of every model are in the Model catalogue.

A population is N units of one model; its state variables are vectors of length N.

using SpikingNeuralNetworks
SNN.@load_units
E = SNN.AdEx(N = 100, param = SNN.AdExParameter(b = 0pA),
             synapse = SNN.SingleExpSynapse(), spike = SNN.PostSpike(τabs = 2ms))
I = SNN.IF(N = 25, param = SNN.IFParameter(τm = 10ms, R = 0.06), name = "I")
E.v[1:5], I.N

(Float32[-54.134026, -56.962227, -63.385036, -59.933155, -55.45753], 25)

Every population has id, name, param, N and records; spiking populations have fire::Vector{Bool}, the spikes of the last step. Most neuron models have v (mV) and an external current I (pA). Any vector field can be recorded with monitor!.

Integrate-and-fire populations

IF, AdEx, Tripod and BallAndStick combine three independent objects:

• param: the neuron model (IFParameter, AdExParameter, ...);
• synapse: how input becomes a current, any synapse model (DoubleExpSynapse() by default);
• spike: reset, refractory period and adaptive threshold (PostSpike).

A step integrates the synaptic variables, computes the synaptic current, then integrates the membrane, detects spikes and resets. For IF (forward Euler):

\[\tau_m \frac{dv}{dt} = -(v - E_l) + R\,(I - w) - R\, I_{syn}.\]

Population(param; kwargs...) builds the population that matches a parameter object, e.g. Population(AdExParameter(); synapse = DoubleExpSynapse(), N = 100) returns an AdEx.

Synaptic targets

Connections and spiking stimuli write into the variable named by their target symbol:

Multicompartment models take the compartment as an extra argument (:s, :d, :d1, :d2; see Multicompartment neurons).

Choosing a model

Populations are Float32 unless built with VFT = Vector{Float64} (see Running simulations). New models: Model extensions.