Spike propagation by synchronisation and vibrational resonance in a feedforwards Izhikevich neural network

Abstract

Multilayer feed forward neuron networks (FFNs) are the basis of various machine learning approaches, in which the propagation of neural firing rates with respect to synchronisation and vibrational resonance (VR) under white Gaussian noise and high-frequency stimulation (HFS) is important. In this study, the influences of HFS and noise on the propagation of the synchronous firing rate and VR are investigated in different kinds of Izhikevich FFNs. For the ten-layer excitatory Izhikevich neuron network, it is shown that synchronous firing rates appear gradually, and diverse noise intensities, synaptic weights and time constants affect the propagation of synchronous discharge rates. For a four-layer excitatory FFN, it is observed that the systemic output rates remain smaller than zero and carry no information on the weak signal when HFS is small. The VR phenomenon occurs when the input of the weak low frequency signal (LFS) and signal output maintain good phase synchronisation, and the LFS is amplified by increasing the amplitude of HFS. In the excitatory-inhibition multilayer FFN, propagation by synchronous firing rates is not good, and few inhibitory neurons remain excited. In the systemic output, the synchronisation phenomenon can be observed but is not as good as that in the excitatory FFN.