Simplified Spike-timing Dependent Plasticity Learning Rule of Spiking Neural Networks for Unsupervised Clustering

Spike-Timing Dependent Plasticity (STDP) is widely believed as a family of learning mechanisms originally postulated in the context of artificial machine learning algorithms, and has been proven successful in training spiking neural networks (SNN). This paper presents a simplified model of STDP learning rule in artificial network based on spiking neurons for unsupervised clustering. Population encoder algorithm is used to convert the continuous real-number data to discrete input spike times. Subsequently, the generated spikes are submitted into a two-layered structure SNN which employed Leaky Integrate and Fire (LIF) neurons as nodes and trained by simplified STDP learning rule. The learning and clustering capabilities of simplified STDP were investigated by applying it to the nonlinearly separable data like the 2-Dimension ring data set and 4-Dimension Fisher iris plant classification problem. The results show that the SNN algorithm can achieve the best accuracy of 100% and 95.33% respectively, which demonstrates that the proposed simplified model of STDP learning rule is a viable training algorithm and the SNN approach is suitable for unsupervised clustering.