Figure 1.

Schematic illustration of the learning rule. For simplicity, we illustrate only the relation between ${\langle {\psi }_i^t{x}_j^t\rangle }_{\tau }$, γtk and wtij, omitting the other components of the learning rule. On the first two lines, spikes of the i-th and the j-th neurons are depicted. On the third line, ${\langle {\psi }_i^t{x}_j^t\rangle }_{\tau }$ decays with time constant τ, and is modulated by the interactions between the spiking activity of the i-th and the j-th neurons. When the firing of the presynaptic j-th neuron precedes the firing of the postsynaptic i-th neuron, ${\langle {\psi }_i^t{x}_j^t\rangle }_{\tau }$ increases (red line). If the postsynaptic neuron fails to fire, ${\langle {\psi }_i^t{x}_j^t\rangle }_{\tau }$ decreases (blue line). The change of the synaptic weight wij at time t0 is proportional to the product of the amount of global signal γt0k and ${\langle {\psi }_i^{t_0}{x}_j^{t_0}\rangle }_{\tau }$ as depicted on the fifth line.