A Novel Adaptive ε-Constrained Method for Constrained Problem

Abstract

Training spiking neurons to output desired spike train is a fundamental research in spiking neural networks. The current article pro-poses a novel and efficient supervised learning algorithm for spiking neu-rons. We divide the running time of spiking neurons into two classes: desired output time and not desired output time. Our learning method makes the membrane potential equal to threshold at desired output time, and makes the membrane potential lower than threshold at not desired output time. For efficiency, at not desired output time, we just calculate the membrane potential at some special time points where the spiking neuron is most likely to output a wrong spike. The experimental results show that the learning performance of the proposed method is better than the existing methods in accuracy and efficiency.