Self-Organized Spiking Neural Network Model for Data Clustering

In recent modern era of neural networks technology, a model called Spiking Neural Network (SNN) was born. This SNN was classified by Maass [1] as the third generation of neural networks. It is a new kind of neural network which is inspired and motivated by the biological neurons ways of communication. The biological neurons communicate with each other through the media of action potentials, often called pulses or spikes. This is a well-known aspect of real neurons, which transmit information by voltage pulses of the membrane potential. SNN has been much considered in attempt to achieve a more biologically inspired artificial neural network. The objective of the SNN as the name implies, tries to overcome the over simplification of the ANN system and emulate the pulse system to come out with a more biologically realistic neural system. This paper presents a preliminary investigation and development of the unsupervised learning processes in SNN. This learning mechanism is extended to one of the most successful paradigm of unsupervised learning: the Kohonen Self-Organizing Maps [3]. Hence, a Self-Organized SNN based on Spike Respond Model (SRM) is constructed. Spiking neurons with delays to encode the information is suggested. Thus, each output node will produce a different timing which enables competitive learning. The model is designed and programmed in MATLAB environment. Further simulation analysis was performed to observe the network self-organizing learning and its topology preservation behavior. The model is further assessed with real-world dataset for data clustering simulation.