Fig. 7: Neuronal diversity fosters complex temporal attention windows in GLE networks.

In the simplest case, a single input signal I(t) is fed into a GLE network and all neurons in the bottom layer have access to the same information stream. However, the output of each neuron generates a temporal shift, depending on its time constants τ^m and τ^r (as highlighted by the neuron colors). Different chains of such neurons thus provide neurons in higher levels of the hierarchy with a set of attention windows across the past input activity. Synaptic and neuronal adaptation shape the nature of these temporal receptive fields (TRFs). For multidimensional input, neuron populations (gray) encode the additional spatial dimension and neuronal receptive fields become spatio-temporal (STRFs).