Time Window, Spike Time and Threshold Boundary for Spiking Neural Network Applications

Assigning threshold value plays an important role in the temporal coding Spiking Neural Network (SNN) as it determines when the neuron should fire, the time window parameter plays a significant role in the SNN performance. This study does two things: First it proposes a mathematical method to find out the threshold boundary in the temporal coding SNN models and second it outlines the input time window boundary which leads to specify the spike time boundary. The latter was used at the former. The threshold boundary method was applied to two learning algorithms i.e., Spiking-Learning Vector Quantization (S_LVQ) and Self-Organizing Weight Adaption for SNN (SOWA_SNN), for both classification and clustering pattern recognition applications, respectively. This method finds the threshold boundary mathematically in both learning models above and observes that the minimum and maximum value of the threshold does not depend on the time input window, time coding or delay parameters in SNN. With regard to the input time window, it finds that specification beyond the parameter boundary affects the computational network cost and performance; also it finds that the delay and the time coding parameters play a significant role in assigning the time window boundary.