Figure 4.

Double population scenario: network architecture, activity and connectivity of the output populations with full (U, τ_rec, τ_facil, A) learning scheme (Part 1). (A) Architecture. The previous network is doubled so that there are now four populations: two input regions (blue) and two output regions ℘^out₁, ℘^out₂ (red). The four populations are organized in two branches, one required to fire at high rates (30 Hz) and the second at low rates (5 Hz). Within each branch connections are all to all (black arrows) whereas initially weak connections (gray arrows) are present between the two output populations and between the two input populations. Input neurons receive a wave-like stimulus from outside (blue dashed arrows). All synapses obey both Spike-Timing Dependent Plasticity and rate-dependent Short-Term Plasticity. (B) Mean firing rate of the output populations, black line for ℘^out₁, and gray line for ℘^out₂. Shaded area represents standard deviation and horizontal dotted gray lines show the two target firing rates (30 Hz for ℘^out₁, 5 Hz for ℘^out₂). (C) Symmetry measure applied on the connectivity of the output population. Color legend as in (B). Connectivity evolves differently in the two populations, leading to a bidirectional motif in ℘^out₁ and to a unidirectional motif in ℘^out₂.