Study of the Influence of Macroparameters of a Spiking Neural Network on the Quality of Image Recognition

In this work, we conducted a study on the influence of macroparameters of a spiking neural network model with classical synaptic plasticity on the spatiotemporal coding of images for a recognition task. The new image feature space represents the average number of spikes as a measure of the neural network's response to a sensory stimulus. Spike timing-dependent synaptic plasticity (STDP) is considered a form of synaptic plasticity. Using the MNIST database, we studied the influence of model macroparameters on recognition performance. The macroparameters are the number of neurons in the model layers and the initial values of synaptic weights in the layer with STDP. The results revealed the values of macroparameters that provide better image recognition.