Quantum-inspired particle swarm optimisation for integrated feature and parameter optimisation of evolving spiking neural networks

Abstract

The paper deals with feature (variable) and model parameter optimisation utilising a proposed dynamic quantum–inspired particle swarm optimisation method. In this method the features of the model are represented probabilistically as a quantum bit vector and the model parameter values – as real numbers. The principle of quantum superposition is used to accelerate the search for an optimal set of features, that combined through co-evolution with a set of optimised parameter values, will result in an optimal model. The paper applies the method to the problem of feature and parameter optimisation of evolving spiking neural network models. A swarm of particles is used to find the classification model with the best accuracy for a given classification task. The method is illustrated on a bench mark classification problem. The proposed method results in the design of faster and more accurate classification models than the ones optimised with the use of standard evolutionary optimisation algorithms.