Innovative Production Machines and Systems Conference

Spiking neural networks (SNNs) are being utilised for solving wide range of problems due to their superior computational power. The connections in SNNs are associated either with a weight values as in the case of a conventional Multi Layer Perceptron (MLP) or with a delay value similar to the Time Delay Neural Network (TDNN). In this paper, two self-organising spiking neural network learning models, namely Self-Organising Weight-Adaptation Spiking Neural Network (SOWA_SNN) and Self-Organising Delay-Adaptation Spiking Neural Network (SODA_SNN) are described. It was shown that these models cluster data sets efficiently and with high accuracy. Their performance was also showed to be comparable or better than conventional self organising neural networks commonly known as the Kohonen Self-Organising Maps (SOMs).

The application considered in this paper is that of control chart pattern recognition. The simulated data set was clustered by the two SNN models and the results were analysed. The performance of the two SNN models were compared with the KSOM and found to be better.