Figure 2.

Structure of a generalized functional additive model (GFAM) of spike trains. In this model, the spike train inputs (x) are convolved with a set of basis functions (b), multiplied with the corresponding coefficients (c), summed, and fed into a link function (g) to form the firing probability intensity (θ). Output spike train y is considered a realization of θ. Note that the output spike train can also be included as an input to model the autoregressive dynamics of the output neuron. The shaded blocks represent the model coefficients to be estimated.