Figure 2 | Scientific Reports

Figure 2

From: E-I balance emerges naturally from continuous Hebbian learning in autonomous neural networks

Figure 2

Selected activities and the average input current for an autonomous SOPBN containing 320 (80) excitatory (inhibitory) rate-encoding neurons, as defined by (1), (2), (3) and (4). The target activity is y t  = 0.2. The time interval is two seconds and the system is shown for the first two seconds after initialization (left), and for two seconds after a previous evolution of 3600 seconds of simulated biological time (right). Shown are the activities of three random excitatory (red) and inhibitory (blue) neurons, together with the averaged E-I balance. The E-I balance is given here in terms of the network-average of excitatory \(\langle {x}_{i}^{(exc)}\rangle \) and inhibitory \(\langle {x}_{i}^{(inh)}\rangle \) inputs (red and blue curves), as defined by (5). The sum (green) is substantially smaller in magnitude for both time intervals. While the neural activities show a large spectrum of activities reaching from nearly completely silent to almost constantly firing neurons at the beginning, the averaged activity after 1 h shows irregular bursts which are characterized by asynchronous neural activities. Note that the large fluctuations in the inputs of the rate-encoding neurons making up the network induce ‘spike like’ activities.

Back to article page