Improving Communication by Resonance across Neuronal Networks: The effect of incorporating resonance feedback
A reliable propagation of spiking activity across weakly and sparsely connected neuronal networks is crucial for the brain function. Usually feedforward networks (FFN), with or without recurrent connectivity within each layer, are used as a model to understand the propagation of spiking activity. Such models have revealed conditions under which synchronous spike volleys (pulse-packets: PP) can be propagated. Two similar mechanisms have been proposed to allow for the propagation of weak and asynchronous PPs. The “communication through coherence” (CTC) mechanism requires that the sender and receiver networks oscillate at the same frequency and phase while “communication through resonance” (CTR) requires that the sender and receiver networks exhibit the same resonance frequencies. Recently, it has been shown that feedback connections between the layers of a FFN can allow for the propagation of weak and asynchronous PPs provided the feedback connection delay matches with the temporal precision of the PP. This mechanism while precludes the need for network resonance and coherent oscillations, increases the connectivity and thereby wiring cost.
In this study, we showed that it is sufficient to have excitatory feedback connections between only one pair of layers in an otherwise feedforward network to reliably transmit weak and asynchronous PPs. We found that the stable transmission of PPs depends on the feedback delay. Two ranges of feedback delays support the stable propagation of the PPs. (1) When the feedback delays are smaller than the temporal precision of the PP, feedback excitation re-ignites the PP before it diminishes to boost the propagation. However, this propagation is fulfilled only for larger populations of projecting neurons. (2) When the feedback delays are matched with the period of network resonance frequency, the pair of layers connected by feedback form a “resonance pair” and locally amplify the weak PP to enable a stable propagation. Thus, we demonstrate that a small modification in the FFN (i.e. few feedback excitatory connections between a pair of layers) can enable propagation of weak signals through a weakly connected FFN without any fine tuning to obtain coherent oscillations or identical resonance frequency in each layer of the FFN.